International Lecture Series 1: Collaboration between UTHM and University of Huddersfield

all right morning Prof garrett can you hear me hi yes I can yeah yeah yeah how are you um I’m Karthi i will be the moderator for this today’s session uh Shall we start the session while waiting to others to come in it might take a minute or two is it okay for you yeah yeah that’s fine yeah okay so uh it’s a fine afternoon here in Malaysia it’s 34 p.m and your side I believe is 8:04 a.m correct yeah mother day right okay then so um a very good evening to all in Malaysia and a very good morning to all in UK it’s 8 a.m there as I mentioned earlier uh welcome to the online session on this uh international lecture series one uh whereby is a strategic collaboration between UTM to be more specific faculty of engineering technology in UTHM and of course University of Huddersfield United Kingdom this lecture series will feature distinguished speakers sharing insights into the latest advancement and innovation in the field of rural transport technology okay so my name is Kartigu Nagaraju and I’ll be the moderator for this event this event consists of three presentation from a great speakers right so to name them will be associate professor Dr joano Preseti Joe from UTM uh very fondly known as Dr joe in our fac uh university right his title will be on re-imagining railway technology road mapap okay can I see Dr joe is already in that uh time and the next talk will be on by professor Garrett Tucker I believe I pronounce your name correct yeah okay okay that’s cool university of Hardersville uh professor Garrett will be presenting to us sharing his insights on introduction to this Great Britain railway industry and last but not least our presenter will be Dr mohammad Musin also from University of Adisville his topic of presentation will be on computer vision for railway applications so it’s for each session the participant will be allowed to question right after the end of the session okay so all the participations will be given e certificates right after the end of the session so kindly please stay up to the end of the session whereby the administrative will share the link for you to fill up your attendance and then only insert will be given to you right okay i believe all our speakers are ready so without further ado let me invite and introduce our first speaker and presenter for today’s event is Dr jawan okay before I I like to give a bit of introduction on a bit on Dr joano Petti Joe dr joano Pascio is uh graduated from RU University Bum definitely I pronounce it wrongly it’s Germany maybe he can tell us later the exact name of the university he had completed his master’s and postgraduate diploma engineering at the Ihe Del University of Technology the Netherlands his main educational research background are on road quality services and also on road safety as as well as track work engineering and maintenance dr joano is another member of working committee in Malaysia under the public transport expert group on the topic I believe road congestion and the ministry of transport Malaysia since 2023 up to now and Dr joano also is a member of evaluation panel ABP at the Malaysian qualification agency academic agency they will say since the year 2018 so without further ado we would like to invite Dr joe Wano which will know him as Dr joe in UTHM to give us a presentation on the title reimagining railway technology road map dr joe the floor is yours okay so hi uh thank you uh PM Ki our moderator and also thank you for the audience also we have this professor G uh nice to meet you I think this is first thanks okay first time to I think meets professor and also thank you uh for the organizer uh to invite me to present them or sharing knowledge i’m not saying there to to teach or to to to just like the sharing knowledge uh which is based on uh our previous works and cooperation with the minister and the agency of the railway okay let me share the slide can you see the slide can maybe you can enlarge enlarge the slide okay enlarge full screen uh I hope it is moving it’s moving it’s moving okay good so uh good goodness okay thank you so again this is what I choose to be the presentation big so it might not too technical like maybe prop Garrett or Dr zen is more just like uh the concepts of uh developments of the road maps of the railway industry and technology which is we are involved and uh sometimes again okay so this is why just like share our previous works and discussion and working group with the the agency and the minister so this part of the presentation today uh I may skip some of them which is uh maybe can be too long so I just to make it as short and briefly as possible so this is I think is it is quite common uh this is also the fundamentals of the background of the our further developments of the rail uh road map real technology uh technology road maps in Malaysia is based on this fundamentals that we facing the climate I think this quite common everywhere in the countries urbanization congestion and maybe the energy but but nowadays we quite concern the most of the congestion especially in the the city of theura that they said so I think nowadays the concern mainly and focus of the congestion okay this is part of this uh global or mega trends of the issue in the mobility okay and again it’s also quite common based on the data what we have and the previous uh guidelines or standard that we have there’s probably the population in 20150 become 43 million something something like that and uh the distribution between the cities in Malaysia I think about could be aso could be the the highest okay Joho and so this is just the population distribution uh in uh the future in about two uh 2015 so this is about the prediction and I think the portion I think still relevant with the nowadays in [Music] 202 but the trend is just like this okay uh it might not be the updated uh the figure of the landbased computation in Malaysia because I refer we refer to the previous database and the uh development of the concept that we developed previously in 2011 that this is part of the landbased competition of the current real networks that we have the the common rail or heavy rail a train light rail monor rail and unicular rail and also further is I I updated it for example is a Malaysian railway for the crater that uh the road is about total in 1,700 kilometers and the KTM for railway is common to use the metal gauge with the total is about 130 station approximately and also we have also the what we call the commuter trade so this is about uh 755 kilometers and I think more than that because previously is about 197 kilometers under construction also met gauge the speed is about 130 is somewhere as you see in the figure somewhere in the north south and the east what we call as the promoter ATM or electric train service also under Malaysian railway uh and this is uh is ongoing I think is about to finalize the construction of the railway uh rout we call the east coast railway line is about 140 kilometers but the gates is slightly different with the previous we use the standard gauge 1,435 mm and the speed about 80 to 60 km with the 20 station with this I can just seems like the train is is using the rolling stop for the CRC China CRC and again this is ongoing I think it’s about to finalize the construction for the civil construction I think and now in ongoing the track construction after the finalized the the piling and the the bridge so now is ongoing for the track construction civil construction so this is connected between Dorbaru and Singapore it’s part of the part Daru and also Synap i think it’s a length about four kilometers but this categorized as mass rapid transit so the size is bigger than the I’ll show that bigger than JQ also LRT so this is ongoing think is about to finalize in 20 26 something something like yeah between the Joharu and Sim so I think this is the sketch or the figure that represent the clang valley rail uh transit net again as I said uh the common uh concern nowadays mainly in Daklang because we deal the facing the high uh congestion in certain section of the road which is considered as critical that must be so so this is why uh this must rapid transit uh map or uh system that might help to solve some part of the solution of the congestion in the backlo okay this is the MRP I think in 2010 so I think more I think it’s about more than uh inlam I think we already more this is again is not updated yet sorry I think it’s more than uh 11 or 12 uh including the MRT 2 okay and LRT3 so I think more than 12 uh links or roots in the area of lamb It is uh I think the one which is to line it which is also a part of the lumbar clang clang uh I think 51 kilome length of the section the MRP1 okay so so the product is rolling stock from the cement thing part of the using the cement okay l1 so this is uh I I just use this uh the the maps of the road is to see that uh based on the [Music] M1 line it’s something interesting that [Music] the the station or the road will be connected with the plan of the highspeed Okay this is why it interesting to show this but not I think not the one okay let’s see the photo slide to the MRT2 that is the and uh Sangra so MR2 is a length is uh 52nd uh 52 kilometer it’s about 37 station which is connected across uh crossing with the MR1 route mainly from the sum route Sang and Putraaya i think the main uh the main uh you know as mean government office so I think this one yes so this is M yeah i think one of these links they will connect it with the uh with the high speed train but sorry I forgot to take the the note somewhere so maybe here we have things one of some of the participant also from maybe know that this that the links uh line could be linked with the highspeed bridge so I think this uh uh oncoming the project I think is uh haven’t started yet but it’s ongoing as I think uh it’s already uh finalized the plan and just about to start the the project but I haven’t start yet okay so this is uh what we call I think MR3 can be called as the circling uh circle line uh among the MRT1 and MRT2 so the length is about 40 kilometers with the station and most is underground 32 kilometers is underground so it’s very less in the elevated uh ground on the elevated so mostly is underground it’s about 32 km so uh so [Music] is for the highspeed train which is previously planned from the following port Singapore so is I think is suspended for sometimes uh I’m not sure when it been started but uh I would like to underline here that you can see that uh there’s the three station will be in the Joh highlight it the location of the station that three of the uh ation will be in Joho that is in mobat and so I think this is interesting and I think this is uh the significance that things that to be uh be aware and concern to consider this as uh the challenge or the opportunities especially for the high tax of the technology or the high speed because three of the station will be in Joho so I think from Garrett know that our city is in Joho okay so in is so this is a challenge and think huge opportunity for our student probably to have this take this opportunity instead of the RTX the previous one from Singapore and to Johar so again this is further the updated uh that somewhere in the Talantan island is also quite uh challenging and also the development is ongoing especially in Sawa that’s connected Sawa and Sabah and other cities as a part of Indonesia and Susabasa Sawa Brunai probably and other cities for the further uh for the long plan construction that will be connected with other city of the countries Indonesia Malaysia and Brunai so that’s the part of the introduction so this railway plans and technology is based on the previous and still relevant of the current plan of the master plan previously by uh land public transport agency uh of the APA uh previously it is called as sp so now is a land transport land public transport agency so uh based on the master plan that uh developed by SP previously that the further or the current national or national land public transport master plan is play a role as guide to the guidance vision give the vision and also to determine the target and also monitoring progress monitoring so this role of the national land public transport which is developed by SPAD having the three main roles as a guideline for the future development of the public transport also the vision to drive the vision of the objective of the national national objective also the targets and for this monitoring broadly it depend of the economic transportation so is about mainly of the economic transportation from the development land public transport to transfer to development of the economy so this is why is mainly to achieve this aspiration based on the economic transport uh econ economic transformation okay so the outcome uh must be the mobility liability and economic growth setting this quite comments for any development of the public council okay again based on the spad and as a uh targets uh from the development of land public transport is previously already identify the 12 national key economic again is about economic transport transformation so there’s I I don’t think uh we uh I will we will discuss this in detail for every 12 national keys uh aonic area but you can see there’s a tourism oil education electrical healthare palm oil communication agriculture business inventor lump also financial service and retail here okay so you may refer to this the guideline i think you can easily find this uh uh book the guideline of this economic competition under SP or so so again this is the aspiration based uh of the 2020 previously that to achieve for example 20 uh most level cities top 20 of economic growth so based on the 2010 population is 1.7 million but uh just check that in 2024 it’s about 8.8 millions so this is a plan of the operation aspiration that based on this economic transportation under the development of the land public transport research what to achieve in 201 so of course based on this demographic government invest investment of the current and the future national agenda so again all the change in demographic is government is under the several of the development development agenda previously economic transportation uh transformation program also here the government transport uh transformation program economic transformation program government transformation program and then Malaysian plan previously in 2010 uh 11 to 15 also is just like industrial master plan that is facilitating the logic development and also some other initiative especially nowadays green technology or state development local local or localized initiative okay so again it might be updated that but the figure is uh shows that until the two 2013 So I think start from the MRT one alone it is said that is about 36 million so this is the schematics of the the patterns of the development from time to time from 2013 until 2013 so I think is still relevant i think some of the updated should be updated in the development of this construction because here I think we have MRT1 ML2 I think RTS not yet here that to be updated RT3 yes okay so the fingers still open for nowadays just some additional uh more project is going on or have been done so the figure must be better than this so where are we now so again based on the might the British design and guideline that the investment right government has invested more than 100 million since 100 since 99 for example with the total employ 9,000 workers so this is the concept of where we are now in the context of the investment so this is what we have uh I think more nowadays is must be more than 60 organization in involving the railway industry including the policy regulator asset design operation maintenance i think there’s a common uh for the railway industrial development structure technology this is what we have here nowadays i think this is still relevant and uh need some updated uh information yeah based on the real operation and asset management so I think yeah the asset owner network here and DRL estate link operator KTM rapid rail opera DRL Saba and SA for the design manufacturing so this this is what we try to map the stages from the design operation maintenance the system so based on the design and manufactur so just try to figure out that manufacturer we identify some companies or local manufacturer or assemblers for example SCI PSI and so for the rolling stock signaling track and electrification so you can see that uh again sorry uh when I updated the data until 2024 we only 2010 here that the turnover shows that sometimes ago the increase and the decrease in of the turn of employment from the technicals and corporation increase okay also capabilities okay let’s see further on the design and manufacturing uh and assembly so this is quite interesting that it is try to enhance the local manufacturers or the local based on the local capabilities and talents okay what I say here in the serum serum is a serum is the standards deployment so okay so serum here able to test for product agent so serum is a part of the entity that may test or development or develop the standard so again based on the spirit to develop the local product using the local talents as a localized spirits so this is why ser is a part for playing this role in testing or development the standard for example so now it just try to just produce the serum just produce so under the Malaysian uh MRC okay Malaysian rail development corporation MRC it’s quite new the NDP here but under the ser uh the serim uh that the development the local standard so under the MDRC uh authority that we just produce the standard But we start from the railway track because this part of the product it may the opportunity then possibility to produce the local standard and the development of the standard that in line with this standard okay so this uh I think uh serim is just to produce uh this is just publish the standard as a Malaysian railway industry i think this is the first standard for the railway track for the Malaysia okay so I think serum is finished uh the producing the standard for the local product and local Malaysian standard so now I think now this uh serim is about ongoing to finalize the standard for the rolling stock so hopefully hoping this coming days I think or coming years there will be another standard Malaysia railway industry standard for ROT stock okay so uh in the context of the maintenance repair and overall so we identify some local prayers and also asset that might be relevant to the maintenance and pay and alcohol here the rolling stock signaling track soification so here then we have SCOI PSI cements APM also rapid trail already but it depends on the level of the maintenance level one two three and also level for component maintenance proposition maintenance modification so this is part of the our main or my main main concern here is the development of the resources human resources or human uh development human capability or human resources or human capability development as a main concern of this my presentation as a part of the the development of uh capability students or localized local talents that’s a student of as part of the local talents the development of the human resources so this is why it’s a part of you can see that UTSM is a part of this human resources development so this is uh this is approved by Mike Malaysian industry governments group for high technology so you can see that UTSM is a already play roles at the earlier I think 201 or 12 so they start play a role some part of the development of the human research in railway technology okay so this is the part of the opportunity for the human capital railway operation in track product design engineer system so you can see the this maintenance and safety okay so as part of because we are play roles a part of the the academic matters or entity so we should not uh far away from the the policy for example under the ministry of educ high education or ministry of human resources for Malaysian qualification agency this is why it is the policy and standard certification is supported by this entity as well because we are play role support on the academic uh side so there must be ministry of education and Malaysia politication agency is a part of this role uh playing the role okay this is uh our problems i think issue are quite common in the skill of the workforce again again UTSM is a part of this development of the skill because we are concerned of the student the graduate the graduate and ability is a part of this role to be the skilled workforce so this is what we are facing quite some times ago i think nowadays still that the maybe the because the attractiveness of the industry less absent of the rail cost so this is why later on from this slide we are one of the the first that the development develop the program of the rail why because this is part of the issue of the problems related cost offer by local absent from the rail related cost offer by local okay so few institution developer railway industry most are conducted in house teaching the industry rail and migration to other industry abroad during specially for shortterm project only requirement for the mult multi- discipline is basically to keep with the latest technology so this is previously defined sometimes ago that become the issue so this is why uh also the same again play role certainly in 2011 or 12 so again issue of the dependency of the import and the foreign technology so this is the main our main concern where where we facing the co last so we feel that this is the critical things and the significance to do it and accelerate that’s the tendency must not be hype for example so the what we have experience during the covid-19 that uh country is a shutdown so it’s difficult especially uh within uh for the maintenance for some maintenance that we are facing that replacing the components or the parts of the railway as a part of the maintenance so there’s a lots of problems for the maintaining during the co so the we learn from this situation so this is the part of the also the problems okay so where we want to go now okay again of course we want to be high and expand dr joe yes uh just a quick reminder we got about 15 more minutes to go thank you what time uh how many more we have we have 15 16 minutes we thought like to stop at 4:00 so that we have some quiz sessions all right okay thank you so high and thank you uh moderator uh high expanding continue to pose the challenge population growth is still become the challenge uh organization policy green technology and also again dependency on the technology and also for the relability safety quality and product so this is what we have to go what to be considered again based on the vision again from the mic that 2030 is become rail industry rail rail industry will be a strong and sustain in the context of the business and satisfy the demand so how we got there so briefly there’s the three goals that identify first ecosystem sustainability and global regional and global so these are the main the goals for the strategic in to support uh that where we go so this is I think key strategies in details but I don’t think we have time to to discuss in details that goal one conductive rail ecosystem then high localization this is the main concern again for goal two high localization of rail related product and also uh goal three is a regional player having the significant market so again this is the opportunity for the b uh design and manufacturer so this may be the opportunity within the context of academic and university that we see the opportunity based on this design and manufacturing of the rail for in the context of the maintenance and overall again it might be the opportunity for the student that will represented in the our program study or syllabus as a part of the not only solving uh the issue national issue or transforming the economic and the growth the economic also in the context of the capability of human resources the development of the human capability of the human resources so this might be identified so we try here to identify the part what the technology of the real so again as I said this is part of the concern my concern here or concern here that based on the real industry stakeholder that you can see also UTSM again based on this map that develop guideline that UTSM is also part play key roles here in the development of the human capability okay for the short and uh shortterm or medium to towards you can see that uh for the shortterm maybe there’s a bridging programs for the graduate student university or college people involving in the railway industry it might be the bridging system training for example but towards 2030 maybe the direct employee by real invest that is why one of the the ideas that this we create another or new it’s not really new I think start on the 2019 so this is why we developed the program that they with they there is the one year workbased learning program so I think this is part of the spirit towards 2030 that might be the student can direct employed by the railway industry because we already take them in the industry in one year so this is what let’s say just like register rather than bridging later on we see the uh later slide okay so this is the current uh highway uh human capability development so maybe previously is unqualified personnel low income qualification established during the job or training based on the type and so on but new or the further might be uh there will be the platform this is called the real center of excellence it might be in the ministry or it might be in the university okay then they create the basic qualifications based on academic program or the center of excellence that might have the training program or the preparation of the student that directly can be employed in the railway industry so this is I think uh some of my friend and colleagues know this or maybe of Garrett already know that we are our our n area of the university is the sustainability technology then transportation okay so railway is a part of the our concern railway because this is our niche for car so again this is why we are having the degree and the master rail in the university program so this is again this is represent for the uh implementation of what we have discussed previously with the ministry or with the industrial pre-industry players so this is the implementation starting 2012 i think we start with the master first in 2012 and the degree start in 2015 so I think just to show you is already since uh 2018 so this is part of the offset or the industry collaboration program ICP there’s still uh many thing to uh to discuss maybe in the concept uh in the context of ICP program so this is what we have for program for the degree program so this is the academic or research development so instead of having the program we also have the research center so I think previously we have quite uh tight cooperation with the Sheffield Bingham I think no so maybe in the future with the other field why not to be part of the our partner in developing the academic or research in the railway technology so as I said instead of having the program degree master so we have also the center of excellence this is part of the center that might provide the training or course on the preparation for the student and graduate to have more knowledge or better knowledge in the rail technology before they graduate or along the graduation so uh I think this is briefly there’s a part because instead of the academic matter we have the center of excellence CO so COE we have our test and responsibility so as a COE center of excellence so we are part of the uh target for recipient for the ICP program under ministry of finance the industry operation so because the center of excellence is on the task on responsibility so then we have it become as a uh main target okay we have uh become the uh one of the target to be the industrial offset industry program so this is why we got here IP9 from the MRP2 automatic pay collection system we have also Eden simulator that really we got from the ICP program under Ministry of Finance because we we are one of the recipient of the ICP program so again we have quite uh cooperation tight cooperation with the MRT so we have some facility in the MRT former MRT uh depot or uh gallery so this is uh industry city uh this is not updated yet so we have some MOU MOA with the industry also the university especially Birmingham so the conclusion is I think the transportation infrastructure and investment is of course to be the elements main element for the economic transformation but this is also the momentum or the quick wins to fundamental requirement for industry for the development of the capability locals localized and also the last is the linkage age of the university and industry could be the fundamental pillar with regard to the railway industry i think uh in this context it is not only let’s say industry that helping the uh university but also university can also helping the industry there will be the mutual uh understanding between industry and the new so I think that’s all thank you moderator and thank you my friends and colleagues to listening for sharing knowledge I it might be not too technical just like profit maybe or profess dr so this is just like the more sharing in the development in the context of academic methods that might be re remap uh re-imaging auto uh changing the map of the railway thank you all right thank you thank you very much Joe so appreciate the portful presentation full of information lot of information you shared regarding our Malaysian railway scenarios okay that’s what’s happening in SP so we could share with our UK counterparts on what is our railway program and our what we call and uh how it works and we mention on the industry and all stuff so very well received thank you very much quite lengthy but on time very nice thank you okay now uh we got a total participants of 75 so I think it’s time for us to open for QA session so if anyone got any questions um can you please uh uh yeah just on your mic and can ask Joe if you got any thank you you can introduce your name first and where you and then we can address your question by the way I think we got some industry people as well i mean industry have join our session today so can we have two questions to Dr sure okay uh I have a question okay okay uh Dr joe thank you for your presentations uh regarding uh the railway stations in Malaysia so uh I just need to ask you um what do you think about the highspeed train in Malaysia or I think Indonesia or Asian uh is it um do you think that uh we need to uh focus on that high speed train compared to the uh right now railway right now moderates do you think is it uh compatible in Malaysia or Indonesia we focus on that uh high speed highspeed train issues okay thank you Dr joe okay thank you uh professor Norini for the very nice uh question because you know that I remember when the project is suspended I think 2018 I think so I remember that the Bernam news interview me about this uh it’s about let’s say about the the same question so uh Profani uh based on my views and experience I don’t think we need the speed now I think the most important now for us is the punctuality or on time okayual it for now for us nowadays is uh relevant and still important nowadays and we don’t need the speed right that’s it why uh and then the question is uh is it comfortable okay from raining I think uh [Music] uh sooner or later the high speed will be there sooner or later but now maybe not yet necessary so why because based on some project is ongoing and project has been done for example gamers Joharu double threat and also RTS Joharu Singapore is probably based on my views and uh analysis is so already suitable and enough for for the let’s say the transforming the economic growth of economic or transforming uh delivering the passenger from the cities surrounding this area because why because we already have the mass almost ready the yohubaru double track so it will be the two double thread between south to north and otherwise okay so this is why uh high It maybe now is not so urgent let’s see what the important is again punctuality having this punctuality can we can not necessary the speed okay and then relate to the Indonesia that already the high speeded you know Remy uh sometimes uh I’m thinking and is it happen high speed train Indonesia who travel with the highspeed train Indonesia it just for the tourist for what you say it’s not for the public you mean public local people uh public not for the purpose purpose of the travel for working for uh for schooling it just for the field to feel and to just like entertainment enjoying enjoying so this is the things why because for purpose of the working main purpose for walking for traveling for school and so on is still not necessary in Indonesia so train [Music] I hope to answer your question sorry okay thank you i have uh another I have an opinion like Dr joe I agree what the Dr they just said because of for now I think highspeed rail is not really suitable in Malaysia because our country is not so big so I think a normal speed train is enough but the most important is the punctuality so from the punctuality we can gain the trust from the passenger then when many many people trust in public transport maybe one day we can have the high speed rate from now I think it is not really urgent because our country is is not too big that we need to transfer from other state to other state for a short of time that’s my opinion because uh when I go training when training to uh in China there is a concern about highspeed trade there is many concern about high speed train because of the cost because of the distance and many things that we can cons we should consider about the highspeed trade for from now I I think we not really need the highspeed train okay thank you Dr thank you Dr isora for your opinion and Dr ju thank you any other question from any other participant hi I can just ask one question please please Prof thank you very much Dr joe that was a great presentation and it’s really good to help us understand the Malaysian railways and it’s great to see the big plans and the developments that are happening you mentioned the original Malaysian railways back in history had a very heavy freight focus with industry can you tell us more about freight on the Malaysian railways now is that a big thing at the moment and are the big plans to expand freight carrying as well as passengers okay thank you for the question G okay so maybe I’m not uh I’m not having the familiar or have experience in the freight transport or logistic in the in this context but as far as I know uh you know as I said in the slide that the ECRL east coast uh rail linkings uh as part of the east coast rail linking is developed for constructed for the purpose of the freight transport because it’s connected be uh between the main or the big what you call free port port and also the mainly the port between the east area and the north area in Kola Lumpo east coast is in Tangano or Plant so it is mean no it is mean that also fre transport or logistic also main concern in this context but maybe that’s not that the priority of the development of concept but it’s still also the concern in development of ECRL is a part of the because developer the main one of the purpose also for the freight transport between uh east north and also to south from there thank you thank you for the question you answer this your question all right okay dr please okay can I add so regarding on the freight so basically our government are moving towards the railroad transportation policy whereby uh the governments are looking forward for the all the uh transportations of uh using a road like for example van lorries and everything is actually moving towards to the road to rail so we are looking more on the freight as well so that’s why we develop our east coast rail so our east coast rail is more uh 60% of the services is for freight and 40% is for passenger so but the rest of our rail network is basically for uh for passenger we have a 50/50 only for east coast rail so East Coast Rail the focus is more on the freight transportations because East Coast rail network will connect with the uh with the we have the um airport as well and connect with we call it as surrender uh transport city so surrender is actually uh some kind of a city that we have a TOD everything that connected all those uh rail network so basically our government are looking more on the freight in future so we have started we have started to develop a policy for railroad railroad transportation yep thank you okay i believe there’s no any other question for Dr joe so Dr joe thank you very much for your Thank you thank you time and wonderful presentation i hope later on we can share more whatever you have shared today right so without further ado I’ll call our second speaker for today’s session which is professor G chaka right on the title introduction of to the GB industry before that I just give some introduction uh on prof G so professor Garrett is the associate director of data science and automation and the institute of railway research IRR in the University of Hardisville and is a professor of railway system engineering prof garrett has over 20 years of experience in rail having previously worked in London in the ground interf technology network rail and RSSB he has a background in mechanical engineering railway systems engineering asset management and maintenance vehicle track instruction and standards development prof garrett leads the IRR smart rolling stock maintenance team which carries out research and consultancy work with the aim of improving rolling stock reliability and maintenance efficiency through the use of robotic maintenance computer vision for advanced condition monitoring and optimized depo workflow planning right without further ado I would like to call on Prof garrett for his presentation prof garrett the floor is yours thank you very much i’ll just share my screen okay good thank you hang on sorry just slight technical issue no it is okay hopefully that’s okay in the slide presentation mode very fine anyways yeah thank you great thank you very much for the introduction and thank you also to the previous speaker so I thought as this is our first one of these seminars I’ll tell you a little bit about the British rail industry and to give you some background and context of where we’re coming from what our priorities and experience are in this country um you’ll see some parallels with Dr joe’s presentation this is going to be 20 minutes and then there’s a more technical presentation from my colleague Dr hussein which will follow on after this so just quickly I’m going to tell you something about the Institute of Railway Research or the IR so this is where I work i’ve been working here for 9 years but the institute itself is older it was established in 2012 so we have 35 staff here we’re full-time research and consultancy in the field of railway engineering and operations we work in a range of areas which are shown in the top left which you can read out i’m not going to read them out but um my previous background is vehicle track dynamics which is what I did my PhD and my early career in i’m now expanding more broadly into maintenance and asset management and modern technologies and you know we might have some future seminars which I can talk more detail about my work um we have four big test laboratories here one is a roller rig for testing train suspension train curving train braking and adhesion one is a train simulator for doing uh research into ride comfort again linked to train suspensions seat design internal layouts one is a panagramraph and overhead line dynamics interaction rig and one is a robotics and automation test lab for train maintenance um I’m going to share these slides with you later if anyone’s interested in and there’s a link at the bottom where you can find out more about the institute and take a 3D tour of our labs and a walk around on the computer so just to give you a bit of context about the great the British mainline network so on the left there just to give you an idea about the country so we’re a population of 70 million people and just to give you some context the GDP is3 trillion pounds which is equivalent to 17.5 trillion reit if I’m pronouncing that correctly um you can see where the main population centers are can you see my cursor moving around so London is the biggest city and the capital birmingham is our second city manchester Huddersfield is in the middle here over on the right top northeast is um Newcastle and in Scotland Edinburgh Glasgow are the two main cities capital of Wales is in the bottom the slight red dot and that’s a smaller city so overall there’s 31,000 track kilometers we carry 1.6 billion passenger journeys a year 16 billion freight tank kilometers um of the network 25% of it is electrified with overhead line and 13% is electrified with third rail so in total you can see 38% of the the track is electrified but that is the most intensively used bits of the network and that carries 70% of passengers in total 10% of freight is carried by electric trains it’s difficult for freight to be operated more freight to be operated on electric trains because freight has to go anywhere including all these routes which don’t have electrification so it’s more flexible for freight operators to use diesel power at the moment we also have metros and trams in Britain so you can see there the red dots we’ve got one two seven trams in different cities so you can see it’s those bigger cities I mentioned that have them and we have four cities that have metro systems i just focused in there on London like I said it’s the capital that’s a big city by UK standards or British standards got 10 million population um in total that’s the oldest also the oldest um metro system in the world opened in 1863 first bit was at the top between Paddington and Farington um now in total that whole network that map appreciate there’s a lot there and it’s hard to see in detail but that map includes some urban rail suburban urban rail as well as metro and trams and light rail and in total that’s a thousand track kilometers and that carries 4 billion passenger journeys a year which is quite a lot when you consider the whole mainline network carries that 1.6 billion a year obviously if you convert it into passenger kilometers the main lines carrying them further just a quick geography lesson because it is confusing when we talk about the UK British Isles Great Britain um I’m talking about the British Railways or Great GB Great Britain which is England Scotland and Wales and that operates as a unit so the Department for Transport is responsible for transport in Wales England and Scotland we also have a separate con constituent country which is part of the United Kingdom which is Northern Ireland they have their own independent transport and railway operations that we in mainland Britain are not directly involved with also making up the British Isles is another country our neighbors the Republic of Ireland and they also have their own separate rail operations now I’ve given you a geography lesson i’m going to give you a quick history lesson so you might be aware this year we’re actually celebrating 200 years of the first railway being opened and that was opened in Britain in the northeast of England in Stockton and Darlington uh so that first operated in September 1825 carried passengers on this route if you can see on the map it’s called the Stockton and Darlington Railway the darker wider line there you’ll see it goes to Stockland Stockton Darlington and actually carried on to Ethley I guess Stockton to Ethley didn’t have quite the same ring to it as Stockton and Darlington railway um that was opened by a engineer George Stevenson led the development and first operations of that um and him together with his son Robert really did a had a big contribution into the whole railway taking off in in Britain and UK and internationally so the first locomotive in the top right hand there locomotion one that’s what hauled those first trains back in 1825 stevenson’s rocket is a more famous well-known locomotive which actually came along four years later and operated on another railway in the west of the country in Manchester um so they carried passengers at 20 kilometers an hour which at the time was just an amazing difference in speed it was so fast that people found it very hard to understand and comprehend they thought all sorts of things might happen to them traveling at such this high speed of 20 km an hour and obviously now we’ve got trains in this country running at 300 km an hour and internationally even up to 7 400 that’s a picture of the people there carriages don’t look very comfortable do they they were just effectively in what looks very similar to freight wagons however that in 1825 wasn’t actually the first railway the first railway was also in the UK it’s actually in Wales very close to where I grew up near Mura Penadarian Iron Works so this was led by a different engineer Richard Travithic and this was the first railway opened in 1804 which is just amazing isn’t it a long time ago 221 years ago and this was inside an iron works it was a freight railway internal owned by the iron works and it was there for hauling the iron ore down into the the docks to to get to market in like nationally and internationally and that first carried 10 tons of iron spread across five carriages and it did actually carry passengers on this first run as like you know big opening gala and 70 people were spread out in these five carriages and that ran at 10 km an hour which again at the time was the fastest pace almost a horse speed anyway jumping back to today so I mentioned there’s two different types of electrification used in British railways one is the overhead line 25 kV this is quite a standard international um type of overhead line electrification for intercity operations so that’s primarily used on higher speed trains our intercity trains run up to 200 km an hour and we do have a short piece of high-speed track in Britain HS1 which runs at 300 km an hour so the electricity is collected through a panagramraph on the roof of the train as an internal transformer which then typically spreads power around in a lower voltage DC and the recur return current goes through the wheels of the train into the runner rails we also have a lot of third rail which you can see that picture in the middle middle at the bottom there’s an extra rail there next to the running rails on the outside of the running rails that’s a 750 DC rail that power gets picked up by a collector shoe which is shot shown in the bottom right and that’s typically used on commuter services so this isn’t suitable for high-speed operations it’s on trains that run at up to 160 km an hour but in a lot of cases it’s on commuted services where there’s quite short distance between stations so it might even be only running up to 100 km or 120 km an hour this is just a little idea of the mix of different types of train we have so in passenger fleet across the whole country it’s 14,000 individual passenger vehicles making up our mainline fleet so by a vehicle I mean a single car so a train could be made up of you know 10 12 16 vehicles for example so we’ve got diesel and electric mix like I said at the start um I’ve taken this reference of the age of the fleet from railway delivery group document that was actually published in 2016 so this is slightly out of date and since 2016 most of these ones the blue built in 1970s which now been 50 years old they pretty much all retired and even some of these in the red have been replaced in the purple so purple pro privatization privatization of British railways that happened around roughly the turn of the century so roughly just over 25 years ago um we also have a large fleet of freight vehicles which in the whole country 26,000 wagons and around 600 freight locomotives all in them um this is all predicted to grow obviously depending on passenger and freight demand going forward in future and we imagine according to RDG predictions the whole passenger fleet will be over 20,000 vehicles in the next 20 years going from its current base of 14,000 obviously that’s subject to various things that may or may not happen along the way um this is just a little introduction into the current highspeed service we have so this is our channel tunnel rail link so it runs from London to Ashford and then into the Channel Tunnel and onto mainland Europe trains operating on this route can go up to 300 km an hour the high-speed section well Channel Tunnel rail link was originally opened in 1994 trains ran high speed in France through the Channel Tunnel and then had to run at 200 kilometers an hour maximum on the British side since 2007 we’ve had this new high-speed route going from Ashford which goes into London St pancress going through Stratford so in 2012 when we hosted the Olympic Games this was opened you know before then and people were allowed to come from mainland Europe straight into Stratford and straight into the Olympic Games right next to the station um so that runs these international journeys using Euroar trains on the left there the one on the right of the picture the more pointy nose one that’s the original Euro Star trains in operation since 1994 then in 2011 the Seaman’s Varo in Euro Stars have been in operation which is the ones on the left they’re a wider gauge as well so also this highspeed line from Ashford into Stratford that’s got that operates bigger trains wider trains than the normal British GB loading gauge um and the reason we have that smaller loading gauge compared to the rest of Europe is because our network is so much older that’s it’s great and we’re very proud to be the first people to have railways and have that 200 year history but it also means all of our stuff is a lot older and a lot of the tunnels a lot of the bridges are a lot smaller so we have smaller trains than our friends across the water in mainland Europe there are plans underway and in construction for another line HS2 highspeed 2 that’s going to be built in phases so the current phase which is due to be finished in 5 years time runs between London and our second biggest city Birmingham so that’s shown in the oops dark green color sorry dark blue color and currently unfunded but the funds in future are the two lighter blue colors are to form a Y shape so to split off from Birmingham carry on to Manchester which is our third biggest city and on the other Y carry on to Leed which is our fourth biggest city There’s a few different advantages of introducing this highspeed line one is that it’ll reduce passenger travel time between Birmingham and London cutting it by more than half so it’ be then quite easy to live in Birmingham commute into London but it’s also a lot of it’s about relieving capacity on the other line so that’s what these green routes are showing these are currently heavily capacity constrained lines so east coast mainline going from London up to York and Le West Coast mainline go which goes past Birmingham and onto Manchester and up towards Scotland and Midland mainline is the one in the middle they’re heavily constrained by so taking more passengers off that that frees up space on those other lines and those other lines can then also carry more freight this is just to give you an idea about the industry structure so we have kind of partially we had this privatization so until the mid 90s we had our railways entirely owned and operated by the government we had privatization in the late ‘9s and up until now we’ve had this private structure and then it’s slowly going back into public ownership but in different parts to it gets a bit complicated the structure but at the moment we’ve got the department for transport led by the secretary of state for transport they have a link with network rail which is a effectively a part a government-owned company and they maintain the track train operations are largely private companies that have a franchise with the government to operate on particular routes and deliver a service under a service agreement And um the office of road and rail may like it makes sure all standards are followed they kind of like police the rules are being carried out by the train operators and by network rail and also any suppliers and have some strategic role at the bottom is a range of cross- sector organizations is the rail delivery group which is in charge of developing strategic plans and helping train operators work together to improve efficiency um maintain safety share best practice in the middle there rail safety and standards board they write standards and also manage the research funding on behalf of the transport funding which universities can bid into rail accident investigation branch so they it’s in the name they investigate accidents and then we’ve got supply chain and also rail ombbitsman which is um more of a financial regulator this is all about to change and we now have an organization that’s going to come into operation in the next few years called great British railways network rail will effectively become integrated into this new organization GBR and al also the main train operators al also become part of GBR freight will still be private companies which will have a contract with GBR and with their customers open access operators are any private still private companies that can add additional services which they think have a commercial benefit and then you’ll still have these cross- sector organizations won’t dwell on this is just to give you an idea there’s a lot of different organizations involved in operating the railway it’s not just as simple as one big GBR or one network rail there’s all these different organizations involved and they have to work together and you know negotiate collaborate just giving you a quick flavor of CO2 which Dr joe also mentioned sorry about this not so familiar with teams with zoom so looking back since 1990 there’s a big push on reducing CO2 emissions and you can see the darker line there is energy so this is electricity generation in 1990 our main form of electricity generation was coal and over that time coal has been almost completely phased out renewables have been ramped up and you can see the energy industry or the electricity generation industry had done a massive massive improvement in reducing CO2 emissions which is fantastic during that time business and residential has also dropped down in CO2 to emission which is a great contribution transport on the other hand hasn’t really made a great sways into reducing CO2 there’s that big downturn in 2020 which was due to co and people didn’t travel very much so we as a railway and as a transport industry have work to do now to try and reduce these emissions due to transport this is just showing things in a slightly different view looking at our whole greenhouse gas emissions so you can see transport now totally domestic transport makes up a quarter of all CO2 emissions on the right is a breakdown of those se CO2 emissions by transport mode so rail is already committing quite a small amount of these whole emissions it’s actually 1.4% 4% of all transport emissions are due to rail and rail carries in total around 6% of all domestic transport so it’s emitting 1.4% of emissions but carrying 6% of all traffic and within that 7% of all commuter journeys and 10% of all freight so there’s a lot of road transport in use and we can see by these percentages if you just manage to switch you know another few% another 5 10 15 20% of all transport into rail with rail under its current level of efficiency we’d already be making a big contribution to reducing the overall CO2 emissions and that is the government target to try and do that especially with a focus on freight Um also reducing freight from our roads reduces congestion on our roads and that makes roads more efficient because there’s less stopping and starting for people in passenger cars i’m not going to delve into this one in great detail because we need to move on to the technical presentation from Dr hussein but just to give you an idea the rail industry has a rail technical strategy this is developed by the industry with industry priorities so this isn’t from the government this is from rail industry bodies that have come up with this these priorities um so this is looking in a sort of 10 20 year time scale of what technical things we need to do what we need to improve about how the railways operate what technology we need to introduce to make things more efficient more environmentally friendly better for passengers um and there’s these five priorities there which listed at the top brake friendly which is a big drive from the government at the moment low emissions like I said optimized train operations efficient and reliable assets and that includes maintenance and asset management techniques and easy to use for all and this is about making the railways as easy to use for every type of person across the country including people with disabilities um some of this you’ll notice parallels the priorities that we just saw from Dr joe um also work on up upgrading the skills of the workforce including digitally talented workforce understanding more datadriven techniques and and this is ongoing priorities just to give you a little idea about the UK rail research and innovation network so the IRR we’re part of UK ukrine UK rail research and innovation network some of the other universities that in around the country that are part of Ukraine and work in rail are shown there in the dots just going to test my geography here and tell you which all the dots are so at the top there you’ve got Harriet W which is just outside Edinburgh in Scotland next do the top one in England is Huddersfield we have Sheffield close by to us then down further the next one down is Lurra birmingham is there Cranfield Cambridge Southampton Bristol and Swansea these are all universities that are working in railway research there’s a link there which says more about Ukrine so Ukrine is made up of three main centers of excellence so the IR where I work leads the center of excellence in rolling stock we have the center of excellence in digital systems which is led by Birmingham University center of excellence in infrastructure which is led by Southampton and there’s also a separate center of excellence in testing which is led by our industry colleagues so that was a very quick whip through GB rail industry i could spend hours telling you a lot more detail about different topics i just wanted to give you a little flavor in this first presentation so like I said we’ve got a long history of railway operations for over 200 years we’ve got a very established network though the downside of that is that some of it’s quite old especially the structures which can be smaller tunnels and bridges which limit the size of train we can put through them gb railways are going through a time of change and this industry restructure and with that we hope will be further developments of new technology coming in and new ways of working um all that remains to be seen how that’s going to work out we’ve got this technical strategy document document rail technical strategy and you can look at that in more detail from the link I’m going to give you with the shared slides what’s great about that is it’s written by railway engineers it’s not written by civil servants in the government um we think there’s a great opportunity to help reduce overall CO2 emissions in transport by rail and the best best way can we can do that is just by transferring passengers and freight from the more polluting transport modes into rail rail itself doesn’t necessarily is already the most environmentally friendly mode of transport but we can even increase that even further we think research and development and learning is essential for future development the railways and universities got a big part to play in that and like I mentioned rail freight I think is an important step and getting more rail on onto more freight onto rail it’s going to be a great way of helping reducing overall CO2 emissions or maybe more multimodal journeys where it’s easy to transfer between road and rail where you take the main 90% of your distance on rail and the last 10% you transfer onto road for distribution locally um that was my very quick go through GB rail industry thank you very much for listening happy to take questions or I don’t know if we need to move on quickly Mr chairman and you let us know thank you prof so uh participants have any questions for you to ask from profil impressive actually see the ecosystem of railway in thank you happy in future to do more a focus session on one of the particular things I’ve introduced if people are interested as well yeah so anyone any question seems to be many requests oh yeah there’s one in chat box i can read up for you that can you brief more on railway investigation branch by mobile zutin yeah I can tell you just briefly they also got a great website so if you look up RAIB or rail investigation branch um you can find out more but that this is a governmentowned organization um they are independent from the rest of the railways um I don’t know exactly how many people they employ but it’s around 50 people um the main workers are the in the investigators accident investigators so every time we have a small accident or a big accident so a derailment would be a big accident um the accident investigation team will go there and they have expertise from engineers human factors railway operations they’ll go to the accident site they’ll measure anything they’ll record all the incidents that happened up to that stage to cause a derailment they’ll look at the maintenance records for the trains for the track they might do some simulation work which they might commission out to experts in my organization for example um to understand the real root cause of every single accident that happened every single accident going from a derailment right down to um you know an accident inside a station you know involving the infrastructure or operations and then they’ll write a report now those reports are publicly available and published on their website it takes around a year for them to fully investigate every accident so you can see it’s quite a detailed in-depth investigation they will share that report publicly on the website and they will place recommendations on the duty holders that were involved in the accident to change how they do things in future um so we had one example like around 10 years ago we had an accident in Glouester freight train and there was [Music] some problem with the track and also a problem with the design of the train raob then made a recommendation on network rail to change some maintenance practices for the track they made some recommendation on the freight operator to do a modification to that particular vehicle which then also went widely to any other freight operators using similar vehicles and as a result of that they’ll significantly reduce the probability of that an accident like that happening again so the idea is this is constant improvement constant learning from accidents and constantly trying to make things safer and safer hopefully that gives a good flavor yeah yeah quite interesting so uh shall we go for any any other question from anyone uh I have a question okay uh may I know how close the gap between academician and university and uh I need your opinion how to make uh the gap closer as close as possible from industry and academicians so we can work together it’s it can be a challenge to be honest um one thing we are lucky with is that we have in this country the the railway safety and standards board I mentioned RSSB that they have a much broader role than the name would would suggest one is that they fund this research this research program 10 million pounds per year so universities can bid into that but that organization they also collect ideas and priorities from all of their members which are the train operators network rails train manufacturers various suppliers and they have a lot of committees and that’s how they collect information from their members so what we do is we’ll place colleagues from the IR or from other university onto these committees these industry committees so then they’re always having this face-to-face time you know maybe once every two months or once every three months with these industry colleagues and contributing to these discussions of prioritizations and actions and idea development so the universities are there understanding really what needs doing also helping to form the scope or priorities for research are then they’re better placed to develop their own staff to meet the needs of industry and understand what opportunities are coming up for industry funded research to bid into it so that’s one great way on an industry level um but it’s a challenge we also try and network as much as possible with industry colleagues and share what we’re doing go to industry focused conferences as well as the pure academic ones more industry events um we also try and recruit people with industrial experience who come with their own network of colle contacts as well as the people with a pure academic backgrounds um so I can’t really got a single answer but I think it’s all about engagement and understanding needs and really trying to train the staff in in real industry needs you know it’s very hard for young academics coming after a PhD or even whilst doing their PhD to apply their good great theoretical knowledge to the real world applications and we try and give staff that opportunity to maybe go and do short placements with industry colleagues to get a better understanding of the real needs of industry this is a massive topic isn’t it hopefully that gives a little flavor of the sorts of things we try so do you think how how many years it can be built the trust between academic and industry from your experience well I think probably at least five years to really get established so this my this organization I work in has been here for 13 years um but between us we have colleagues who’ve been in working in railways for yeah myself over 20 years or even 30 years um but yeah but over a period of a small number of years I think that can relationship can really grow with focused effort and delivering some key projects which show some tangible results thank you bro thank you thank you is there anything else any questions do you have any question oh yeah sure sure sorry uh I’m uh Joe from Indonesia i am uh under uh graduate student from uh University of uh Jimber uh so I think uh uh I want to compare about the maintenance in the uh uh England and Indonesia because uh the basic maintenance in Indonesia is basically from the track uh uh track uh measurement car track measurement car so it is uh make uh track quality index every 3 months uh I I heard or I I read in in uh I think Dr y is used in uh England because it is comprehensive car i think I think it’s is excellent uh Dr y is excellent it is because it is uh completely one in one uh one in one scale model for the for the transet yeah so uh uh what do you think about the okay the frequency of the yellow uh uh inspection the track all track in England every year to to uh to make a decision for the maintenance thank you for it great yeah so network rail have got a fleet of around 10 different trains that measure different things so the track geometry one which I think was the maybe the focus of the question that has maybe four different trains that run track geometry and they prioritize depending on the type of route so there’s seven different categories of line um the busiest category is the top category is 1 A and then 1 2 3 4 5 6 and depending on the category of line they will inspect more often so a category 1A line will be the 125 mph intercity routes that have a certain amount of um pass train density so the calculation to put it in the category is based on speed but also number of trains per year um but a category one 1A track will have a track geometry inspection once a month so 12 times a year um and then I think category one I can’t remember the exact numbers it’s something like every two months and then down down and the lowest very lowest category which will be a very slow track which doesn’t have many trains on it that’ll be once a year okay and that really helps focus resources so it’s focusing the resources on the track that’s going to change the most often because it’s got the fastest heaviest most frequent trains running on it and also the ones that carry the most risk for the the same reasons uh so in England uh the category is based on speed or based on annual track p annual uh tonet spacing tonets annual pacing it’s an influence of both speed and annual tonnage which one is uh it uh if the one line is most in the passengers so and then uh one line it is for freight train it is uh not uh not not the speed is maybe low but uh the annual passing tone is is high uh so uh which one is make uh uh most important there’s an equation that works it all out but in the speed has a higher waiting than the total tonnage sorry sorry the speed the speed will have a a bigger waiting will have a bigger factor on the track category than the total tonnage will so oh okay you know if you have one route carrying the same tonnage of freight at 75 miles an hour as one carrying the same tonnage of passengers at 125 that will be a significantly higher track category that passenger one that runs at 125 okay okay okay understood okay thank you great so I think uh we are done with the question session of K thank you very much for your wonderful presentation sir so uh to move on further uh shall we call our third presenter for today Dr mohamad Hussein right welcome Dr mo Hussein so just a bit brief introduction so Dr osan received the battle of engineering electrical electronics from an MSE in IoT of things from the university of hardest field in 2019 and also his PhD in AI artificial intelligence for industrial defect identification not sure which was that mention here his research is focused on the detection of industrial fault via computer vision is a particular interest in the field of machine v machine vision focusing on the development of lightweight architecture that can be optimized for development on constraint age devices and ultimately onto the production floor he is also researching into design level architectural interoperability with a focus on explanable AI for sensitive fields such as medicine transportation healthcare is keen on the deliation of vision vision aim at providing tangible solutions to small and mediumsiz firms a bit hard for me to read your brief actually so much complex terms and I think by the way uh Dr moad Usin who will be with us on the title of computer vision for application right sir please the podium is yours thank you thank you thanks for the introduction right uh let me just make sure I get the screen on so just let me know if you can see my screen please okay get down to slideshow hi everyone so uh yeah welcome u and thanks for having me for the presentation uh so I’ll be looking at u I’ll be looking at it from a more technical perspective so again the focus is railway but then there’s a particular aspect of AI which we’re working on within the University of Hersfield and our railway department on how we can integrate computer vision uh into the railways so I’ll be giving an overview of what that is what computer vision is um and how we’ve started doing some work with computer vision and applying it to different aspects of the railway uh to be able to enhance uh procedures operations and overall efficiency so I think just before we go into computer vision it’s important just to give a bigger picture so where does computer vision actually fit in and also you know the landscape at the present times where are we so computer vision essentially is a subset of artificial intelligence within artificial artificial intelligence so AI is really an umbrella term uh it’s a buzz word which is used quite loosely uh what it is is essentially is there’s a few pillars that make up AI one of them is machine learning which is the most generic term then we’ve got deep learning and then within deep learning you can say we have another subset which is known as computer vision so our focus is on computer vision now again coming back to the umbrella term why is AI so important now uh I mean the term was coined back I think around 1956 so so it’s been around for a while there were the first neural network again was in the uh mid 90s so work has been done for a while but why has it suddenly uh seen a lot of surge in its uh importance so in terms of on a national level AI is essentially now a national asset so the reason for that is it’s driving economic growth uh for national security and also innovation so all of these pillars you know AI can be seen that it is leading in all of these areas and that is the primary reason why governments view AI uh as a critical pillar and we can see the manifestation of this by looking at what the governments are actually doing so there’s been key developments worldwide uh here in the UK you know we now have a dedicated national AI department so which looks at what what are our priorities with respect to artificial intelligence what do we need to focus on what’s our road map and then a lot of experts feed into this so we can create our strategic road map uh USA again uh I mean there’s these days there’s executive orders coming out from everywhere in the USA but AI also has executive orders they’re also prioritizing it there’s huge amounts of investment coming in the new administration uh in the US in its early days you know did commit a lot of investment to AI uh EU came out with the first world comprehensive AI regulation so they’ve got their own EU AI act uh and then obviously we can’t miss out China from this china again is a front runner uh and they have a lot they have massive public investment uh within AI and we saw a manifestation of that uh in another subset of artificial intelligence uh where you know they came out with deepseek um and it’s been given a hard time to a lot of the UK set uh sorry the USA setups in the same domain so they also are front runner in this um and then you know our countries like Malaysia the ACL nations so they are also setting up their own national AI offices um because one of the real concerns is the regulations around AI uh a lot of the governments do not want it to go unchecked so uh for that reason it’s they are creating their own departments and they taking in uh inputs uh from the expert to see okay what kind of guard rails can be put into place uh but like I said in general why is all of this important now because the governments are seeing the importance of it they can see what are the benefits they can see what are the advantages and what are the disadvantages and these are not just at a local level you know they at national level and they at a level that if they’re not addressed uh national security can be at risk and we see we’re seeing a lot of this you know within the warfare uh domain at the moment and all you know with the use of drone technologies there’s a lot of AI utilization over there so AI in general I mean the gist of this slide really is just to set up uh just to give some context that AI is at the forefront almost in all of the countries at the moment so okay so then that leads us up into computer vision so like I said computer vision is essentially a subset of artificial intelligence and going deeper into a subset of deep learning now the gist of computer vision is to be able to get machines to see so visualize interpret and understand visual information so usually when we’re looking at machine learning we’re looking at deep learning we’ll have you know a bank of numerical data and that data is given into a machine learning model you know such as a decision tree or a random forest or the artificial neural network um and then that will do some model training go through some back propagation mechanisms uh and come up with some generalization pattern uh and based on that we can infer new results now with computer vision we want to take visuals so we want to take images we want to take videos directly and be able to infer on top of them now there’s a lot of benefits of that so we’re able to essentially look at something the way humans are able to view so based on the visual cortex and then that speeds up a lot of processes and it opens up a lot of domains where going through through the numerical procedure of collecting images doing the conversions doing the manipulation transformations and then getting it into the right format before feeding it in and then repeating the whole process again once you’ve got an output it can be a can be a quite a long-winded process so the visual mechanism of it is to be able to make it as easy as possible and get the computer to essentially see the way humans are able to see now what are the core areas within computer vision so we’ve got image classification so this is basically just looking at an image as a whole and just stating you know either within the image for example there is a cat or there is not a cat so just image classification as the sole purpose we then move a step on uh an enhancement to that is object detection so this is within the image being able to say okay within the image there is a cat but where is the cat located so being able to get the location of the cat again that’s enhancement and it doesn’t just stop there it’s not just a single object it could be multiple objects so you know there could be a cat there could be a dog there could be something else uh and then the CNN model is able to detect all of them understand the context and be able to locate actually where within the image or where within the video frame are these objects located um going on top of that so we have semantic segmentation so this is pixel level so you know again we’re going more granular not just uh looking at the location but looking at the location with respect to the pixels uh and then we’ve got a lot of applica so some genetic application that is facial recognition uh optical character recognition these have been around for some time and these have given the inspiration for us to expand into new areas new branches uh and one of the big branches there is defect detection and that defect detection you know within manufacturing facilities industrial production flows uh and railway comes into that as well so essentially the goal of computer vision is for us to be able to give machines so thus you know computers smart cameras generic cameras the ability to be able to see and by being able to see they are then able to make strategic decisions based on that visual data so let’s start to tailor our discussion now a bit more towards the railway now that we have a bit of a background to it so let’s start off with a bit of a scenario uh now imagine it’s 4:00 a.m so it’s pretty early in the morning way too early for me and you have a a rail inspector and he’s walking miles along a lonely stretch of track so this is early morning uh and if you if you’re in the UK you know most likely it’s it’s really cold uh it’s raining and he’s got a flashlight in his hand and he’s trying to spot cracks loose bolts or weeds which are pushing through the rails so on the railway track uh like I said you know it’s in the UK the weather is cold visibility is very poor around that time and every hour which is missed increases the chance of a delay or worse a derailment so now this kind of setting you know back in the days was a standard so it was a normal procedure so what I mean by that is manual inspection was the standard that we would do for uh you know railway track inspection critical component inspection and to a degree it still is uh but we are moving towards AI so I’ll come on to that now what’s the what’s the disadvantage of of this kind of manual inspection so a it can be slow it is edprone it’s deeply reliant on human focus so human bias will have a big impact on this now there is also AI bias which will come to but human bias is always there and it’s you could say it’s inherited by AI bias and all now what I mean by human bias here is there’s the likes of fatigue you know how was the last evening for that particular person what side of bed did they wake up from how are they feeling so all of these different characteristics um are impulsive in the sense that based on how they are at the at a particular type that will have some kind of influence on their judgment so um if if they’re having a bad day maybe you know they’re going to miss a lot of uh hazards a lot of defects uh cuz the focus just won’t be there so essentially manual inspection led by humans really does depend on the mood of the human or the kind of bias which is inherited by the human at that particular time um and this could be detrimental you know if we’re missing hazards depending on the nature of the hazard depending on the nature of the defects you know there could be millions lost in damage and worst case scenario there could be loss of life so there were a lot of well there’s always been these um these kind of uh backtracks with manual inspection so these kind of loopholes they’ve always been there and we’ve not really been able to address them because we didn’t know what the alternative was um or you could say we knew what the alternative was but we did not have the means to be able to utilize the alternatives so moving ahead now picture the same scenario but instead of that human you know trolling along 4 4:00 a.m in the morning we’ve got a high-speed train which is racing down the track it’s not carrying any passengers uh you know it’s just a standard procedure but it’s been equipped with smart cameras which are hosting artificial intelligence algorithms in particular convolutional neural networks and these in real time are able to flag up you know micro fractures in rails identify overgrown vegetation uh be able to spot missing fasteners um and a lot of other defect and they’re able to do it in real time uh at a very quick pace now that takes away a lot of the limitations which I just mentioned uh a key one which I didn’t mention was the labor cost and all so you know being able to get inspectors on there those costs so not only the buyers but we’ve got the financial cost that comes with it whereas with this kind of scenario where you’ve got smart cameras they don’t have to be on high speeded trains you know you could be utilizing drones you could be uh utilizing your standard trains with for example we’re doing experiments with forward facing cameras on standard trains cutting passengers to see are they able to locate damages on the rail track and be able to report them back in real time so this is a high-end scenario but this could be utilized or generalized to standard trains with forward facing cameras so that essentially is one case of utilizing computer vision within the railway so you’re moving from reactive to predictive maintenance and from slow and subjective to fast accurate and always alert so you know the AI algorithm will not uh will not be rendered to fatigue it will not need breaks you don’t you don’t need to pay the AI algorithm for the work it’s doing at the end of every month so there’s a lot of these disadvantages which come with human inspection which can be done away with when we’re utilizing artificial intelligence and in particular computer vision so what if AI could prevent railway accidents before they actually happened so that’s one of the areas we’re also looking into now just to give a context there so 70% of passenger injuries are actually linked to door component failures so this is with respect to the railway now rail accidents often start small so this could be you know a loose bolt especially when we’re talking about critical components it could be a worn out wheel uh it could be debris uh you know on the tracks uh overgrown vegetation uh you know which has uh no one’s really looked into it or it’s not been noticed and it’s just it just keeps on growing so traditional inspection so where I was mentioning you know human based inspection there and with all of the limitations it comes with uh it can go up to as much as 30% of critical faults being completely missed out so that is a significant number you know even one critical fault which is missed out could have uh you know massive consequences here we’re saying 30% of critical faults based on uh traditional inspections can actually be missed out so there is definitely a use case here for automation through artificial intelligence and computer vision for us to be able to improve enhance and optimize our operations within the railway so computer vision is able to bridge this gap between inspection and prevention now there are challenges so this is a test pi which we have within the university of huddersfield within our railway department and we have started doing some work on okay how can we apply computer vision to be able to let’s start with just detecting the critical components and then based on that we can move on to further inspection of different components uh and then move from a bogy to you know a complete train so I’ll show you I’ll show you how a computer vision deployed onto that actual video frame looks but let me just give you a bit more of a background first so before I show you the output why is so we mentioned why is AI important now why is computer vision also ideal now so computer vision like AI has been around for a while you know from the like I said artificial neural networks and then we had within the 19 1990s uh there was work being done within computer vision but it didn’t really take off until you know 2012 uh or you could say 2015 so 2012 was when the first real computer vision algorithm came out which was known as AlexNet um and then we started to see a lot of advancements cuz it was able to take in images uh it wasn’t very deep it was able to give you good inference uh but it was only until 2016 when we had an architecture known as YOLO so YOLO stands for you only look once so I know that’s also you only live once but over here within the computer vision domain we’re focused on you only look once uh so YOLO v1 came out in 2016 and that was a massive breakthrough uh because YOLO v1 is a single stage detector now before that you could utilize computer vision so you know for example we wanted to apply it to be able to detect critical components on the rail bogy what we would need is you know massive GPUs which are expensive the and the reason for that is because we need a lot of training data set we need to train the model and the model would be trained on architectures such as VGG such as ResNet and these were very very deep architectures they had loads of layers and their training meant that they would require a lot of specialized hardware uh in particular GPUs to be able to do that work so YOLO said okay let’s move away from two-stage you know highly deep architectures to single stage detectors which can do a complete inference within one room and that was a massive advantage because what that meant is we could really scale down on the hardware that was required for people to be able to you know train their architectures and essentially it bought computer vision into academia into you know standard research researchers people at home could start utilizing it they could start training their own models uh and there was a lot of advantages of that you know because of that the advancements were much quicker uh and just to give you a taste of how important YOLO has been or the role is played within the computer vision domain is in 2016 we had the first YOLO version which is known as YOLO V1 and now we’re in 2025 so over this period from 2016 to 2025 you know around 9 10 years we’re already at YOLO V12 so we’ve gone from V1 2 3 4 5 up all the way till yolo V12 in the span of just around 9 to 10 years so there’s massive amounts of work been doing the advancements are very quick and all of the focus with YOLO is how can we create architectures how can we create lightweight artificial neural networks or convolutional neural networks which can be deployed onto the edge do not require massive GPUs and can be utilized for standard real world applications because what normally happens is a lot a lot of the AI is constrained within academic worlds so you’ll have these researchers uh that are sitting down creating new architectures uh publishing papers but they never get onto the production floor you know they never actually make an impact within the real world um or on the other end of the spectrum you’ve got the Teslas and the Googles and the Facebooks um now these companies will create something that does have an impact but they’re able to do it because they have the money and they’re able to have you know massive amounts of GPU leverage which normal people don’t so that’s one of the things that now with the advancements in lightweight architectures we’re able to give that kind of leverage to the layman you know to the normal person and they able to create architectures which is really advancing and creating a lot of new technology within the computer vision domain so okay I’ll move a bit faster uh because I’ve got quite a bit of slides to go through and I’m just looking at the time so essentially computer vision uh we have a lot more efficient architectures now we’re able to deploy on the edge so what that means is we don’t we no longer need you know high-end GPUs for deployment we’re able to get it onto for example a Raspberry Pi i don’t know if you’ve came across a Raspberry Pi or a Jetson Nano so these are onboard edge processing units uh not very expensive uh some of them have onboard GPUs like the Jetsons and you could have a full-on computer vision model deployed onto that uh and close to the source again that has its advantages especially you know when you’re looking at from a cyber security perspective when you’re wanting fast inference when you don’t want to spend too much money you’re low on investment so there’s a lot of advantages there I’ve I’ve touched quite a bit on reduced GPU requirements um and also to facilitate a lot of these computer vision work now we have improved frameworks one of the key standouts is known as ultralytics so this is the one that manages YOLO essentially now but there’s also other ones you know we have Roboflow we have Luxonis so we have a lot of different frameworks which makes computer vision easy and you’re able to do it very fast and there’s a lot of broad applications so in industries are utilizing computer vision now they’re understanding its use cases they’re appreciating the use cases which is you know a good start and we’re also starting to mature with respect to data sets so you know data sets are being built and the re researchers are able to utilize these data data sets to create specific models which can then be deployed back onto the production floor so in terms of I’ve mentioned all of these benefits you know of computer vision but when we look at it from a railway perspective there are still challenges so it’s not as easy as let’s take let’s record a video and then you know put all of that video into an ultritics framework uh train a model and get some results and then you’re ready to go there’s actually a lot of challenges that that are present within the railway domain and in general within the manufacturing facilities now we have occlusions so this is referring to where the target object or the object of interest you know is occluded there’s some kind of overlap with other components and we can’t completely see this component now we can’t completely see it but that is the component we need to detect so we then need to find a way of how can the computer vision algorithm understand that okay that is the object of interest but I don’t have high confidence that it is actually the object I’m looking for so what are the ways around that then we’ve got lighting variations so depending on where you are environmental you know the looks intensities of the light will also play uh can deceive the CNN model so you know if it’s too dark compared to if it’s too light the kind of visual output you’re going to get is going to be very different so how can the model generalize and understand like a human would that actually all that’s happened is the lighting has changed the looks intens looks intensity has changed but that is still the target object I’m looking for and then obviously there’s a massive scale of diversity and what I mean by that you know we’ve got you might be looking at detecting large engines uh compared to very tiny bolts so some of our initial work was not uh done on detecting critical ical bolts and the architect is known as bolt vision which has been published and I’ll I’ll share that with you uh and another key uh challenge we have is limited data so you can get a lot of data in general of standard use cases but to be able to get a lot of the defective samples can be hard so you know when you’re in within uh within a production facility so we’re looking at doing some work uh with the Delway production facility now they can give us a lot of data but when it comes to giving us hazard data so we can train for hazards they have very limited data so how do we overcome that challenge how do we create enough samples of hazard data to be able to train a model because your model will only be as good as the data you give so a key principle in computer science is you know garbage in garbage out so if you give it rubbish data you know your output only what you can expect back is also going to be rubbish so you need to give it the right kind of data and the sufficient amount of data for the model to be able to learn generalize and then give you accurate results okay so now the next slide I promise I will show you an output of how the model actually looks when it’s deployed onto that rail bogy I showed you initially uh let’s just cover this slide before we get on to that cuz the model you’ll see in the next slide which is deployed in real time is based on YOLO yolo v10 so I think it’s just good to give an overview of YOLO um and also why I keep on going on about YOLO because this fundamentally is the key model or the key framework that has changed computer vision you know like I said from 2016 to 2025 you can see how many new architectctors have came out based inspired by YOLO now I won’t go through all of these but there’s a few of them which have really advanced YOLO now this was started from YOLO V4 V5 when we started to move towards industrial application so before that there was a lot of genetic work you know you’ve got open source data sets and you’ve been able to improve the accuracy on that from YOLO v4 v5 onwards we’ve started really focusing on okay how can we address real world applications how can we actually make these architectures lightweight utilizing for example attention mechanisms single stage detectors uh a mix of augmentations to be able to really focus on an architecture that can be deployed within the real world scenarios um and then you know you’ve got the likes of YOLO V6 a key success for YOLO has been that it’s not owned by anyone particular entity you know anyone can create a YOLO architecture and if it performs good you know on the data set on the benchmark data sets you know the Koko MS cocoa data set and the other data sets that are out there then it’s recognized as a real advancement in YOLO so what that means is it’s not tailored or it’s not taken ownership by just one entity you know anyone can be working on it I think YOLO v6 was done by a research uh group in China yolo V4 is done by Ultralytics then YOLO ultritics came back with YOLO V11 uh sorry V10 so there’s a lot of different uh companies entities research groups which are working which are inspired by YOLO and creating specific architectures to forward the agenda of industrial applications okay so this is uh the initial video I showed you uh but with critical components being detection detected through a smartphone by deploying it on a YOLO V10 so you can see here in terms of the inference is pretty fast uh the real reason I wanted to show you this is because this was the first experiment we did uh with the railway department in terms of how we can start looking at you utilizing computer vision now one of the impressive things here is all of this was done within a day so you know from collecting that raw video data which we showed you to splitting it into images labeling the key components then training a YOLO model and deploying the YOLO model back all of that was done in a day whereas if you go back say 10 years you’d be looking at getting something like that done within 6 months so that just shows the speed of advancements and the impressiveness of YOLO in the sense that we can really speed up deployment now if we’re able to detect components there we can then you know that opens a lot of branches cuz in you’ll see in another slide is we then only zoom into the axles so I don’t know if you can see my mouse moving over here but we then focus on a particular component because that particular component can have sub sub areas or sub domains of interest so for example this area here the axle end cap that has bolts on it now what about if a bolt is missing so we then are able to zoom into that component and then focus on the subin interest objects of that component to see if there is anything missing or not so it opens up a lot of different branches but that’s only possible you know due to the fact that we’re able to speed up the process we’re able to do it in real time and get performance back in real time to be able to make strategic decisions yeah so just zooming into that uh well basically what I just said based on an abstract view we can then zoom into certain components within there you know bolts notes washers seals and there’s absolutely lots uh we focus at the moment on rolling stock so there’s a lot of areas within the railway rolling stock domain that can be applied to with respect to computer vision and I’ll show you some use cases there so before I do uh there’s one key challenge which we faced before we created u our first architecture which is known as bolt vision now that was being able to detect smaller components um because even from a human perspective you know the bigger components are quite easy to detect they’re quite straightforward the smaller ones need a bit more focus even from a human uh from a human perspective you do need to have some training to know what you’re actually looking for um and due to that you know uh some statistics show that 25 to 50% of these kind of defects uh when done through manual inspection can be missed so there is a real use case for us to be able to automate the process remove the bias give good training data set go through a standard but effective augmentation pipeline and then train a model which is able to if not eliminate but at least reduce you know the statistics we’re showing over here now so this is one of our uh first architectures like I mentioned B vision so the bigger picture you saw of the complete Borgi this is now focusing on a particular area and that’s the axle tab and being able to see okay are there any bolts missing again you know from this we can clearly see a use case of it because critical components um being missed out and you know that particular bogy going into operation can have you know really bad consequences so to be able to do that now we don’t claim that computer vision just needs to do it on its own it can also in its early stage be led through human inspection so it can be an assistive technology whereas for example a human is doing the inspection but he’s wearing a smart cap which has a AI computer vision camera embedded in it and whilst the human is doing it the smart cap is also doing its own inference and if the human does miss something out you know the computer vision can label that or can just give a prompt to the human and say by the way that there was a missing ball there have you have you seen that have you factored that in have you you know have you noted that so it can also be used as an assistive technology um and then there’s a lot of technicalities I don’t think I’ll have time to go into that with respect to the architecture but also interpretability so we’re able to generate these heat maps which you can see on the bottom uh right hand corner so whilst you’re training the model we now have frameworks we have libraries that let you see what is the actual model looking into you know why is it making a particular decision so for example you know if the model is giving higher weightage to this area then we know this model is false it’s not learning the real pattern it’s actually being falsified or is being misled by this tag which is over here so it’s been misled by this uh label thinking that this is a bolt whereas actually the areas you should focus on are these so these are where the missing bolts are so by being by having you know interpretability we’re then able to have the confidence that actually has our model picked up the right patterns or has it just been is it mclassifying and if you remember you know if you go back 10 uh 10 years or so uh interpretability was a big issue cuz uh artificial neural networks the big issue we have with them is everyone says the black box so you know you’ll put something in you’ll get an answer but where did the answer come from what was the logic behind it what were the justifications we don’t know so there are frameworks now being put into place because not only in manufacturing but in more critical applications you know like healthcare it can it can have a real uh bad consequence if we don’t know where the judgment came from which is why areas like healthcare are very hesitant uh in terms of deploying these kind of models in very critical components but again the frameworks are being put into place there’s new methodologies coming out which are able to show interpretability in terms of how did the model reach its final conclusions yeah so this slide basically just touches um in a bit more detail what I’ve just mentioned um again I’m not going to go into too much detail there cuz uh we are going to be short on time but the ar here’s another kind of architecture which which is known as a vision transformer and we embedded that with a convolutional neural network so we combined two architectures to be able to generate enough samples and to do inference for critical component detection just on the axle end to zooming in on a particular component and again like I said we we were able to factor in in terms of the interpretability through the training process to see if our model was actually learning the real types of patterns or if it was being misled by you know some other component which is on there or a sticker or something which has been placed on there um and then some details of the architecture over here in terms of how we start with the input what are the different procedures the input goes through how is the training done how it goes through the dense layers the dropout and finally it makes a decision you know is the bolt there is it missing or is it present so again there’s a lot of details there but we really don’t have time to go into them uh yeah so like I said in terms of how could this be beneficial you’re looking predictive maintenance scheduling post repair validation so once something is once the validation process has been done you know can we validate it through computer vision safety compliance auditing so there’s a lot of different applications for it and the interpretability factor also you know gives us more confidence to be able to say okay let’s bring in the computer vision into our existing validation processes cuz we’re able to justify how we got to the end means okay now the actual defect of the problem does not just have to be you know with components on the train it could be external to that so another area we were looking at was railway tracks so outside damage you know 20% derailment risk there is there is a lot of hazards that can occur outside of the train in particular with railway tracks so uh some of the areas we’ve been looking into is being able to detect animals so this is not a major issue within the UK but uh we’ve also been looking at some footage from Pakistan from Bangladesh within other open areas you know where animals are around the railway track area and then there is real risk of uh collision there but there’s also other use cases you know vegetation is a big one flooding is another one that needs to be taken into account depending on the demographics fix yeah so I’ve just touched upon this and then another architecture which we were looking into i’ll show you the output of this architecture but was the YOLO V11 so again focusing on YOLO YOLO to be able to detect in real time objects animals humans so here you can see a an actual application which rather than having a massive GPU uh on the cloud and sending data back to be able to do detections we can have a Jetson Nano for example or a Raspberry Pi that’s actually put on top uh on uh sorry on a forward facing camera is hosted within that and that architecture is able to do the inferences in real time to be able to give you an output now you might have noticed is it’s categorizing that as animal hazard whereas it’s a human that’s just a label header so you would change the label when you’re doing the training from a animal to a human and then there’s also other factors that have to be taken into account so over here you can see you know vegetation which is really overgrown uh you’ve got animal hazard there you’ve got flooding which is starting to come onto the track so there’s loads of different types of hazards so a multimod model uh architecture can be deployed depending on what are the needs for that particular area and like I said after mentioning all of the benefits of computer vision but there is a lot of challenges which we still need to account for uh and the biggest challenge really is limited real world hazard data set so we had a company that we worked with and we were looking at being able to detect vegetations on rail tracks now they gave us absolutely lots of footage when it came to uh you know normal standard rail tracks and the train is just moving across but there really wasn’t much samples of the actual hazard so you know overgrown vegetation there wasn’t much hazards so how do we create our own hazards but not do them as an overkill so don’t overextend them but be able to generate real world use cases and you know gen AI plays a big part in that so there’s architectures like stable diffusion if anyone’s came across them which help you to be able to generate new images new samples but then you still have to account for okay are they realistic you know are they representative of the real world so the bottom line really is you always need the number one go to is real data and you train the model on that if you can’t get real data or if you can’t get sufficient amount of real data you need some amount of real data and then you move towards gen AI stable diffusion augmentations transformations to be able to generate new data but always benchmarking with the original data to say okay is this data representative of what I might see in the real world and if it is then that can be used to train the model if it isn’t and you still use it to train the model then you’ll get a model which is not realistic it’s not generalized to the real world application so yeah so this is basically just touching upon how a stable diffusion framework can then be utilized to be able to generate new samples based on images where we don’t actually have sample hazard samples or we don’t have sufficient amount of samples there so I won’t go through the inner workings of that because we’ve got a couple of minutes uh so here are some examples of how you know based on real images on the left hand side we can give it some prompts to a stable diffusion network and we’re able to generate new samples like I said some of them can be an overkill so for example if you wanted to generate flooding you know this is sort of good but this is an overkill so the one at the bottom you know you in most cases you won’t get that so again that is over augmented and then you know based on that again like I said you can utilize gen AI for generating loads of different samples you can get as creative as you want with it but getting as creative uh you still have to understand okay what are the real world cases you know is it still representative otherwise you can be creative but the model you you will get at the end of it will not be useful uh so yeah a lot of our work at the moment is on how can we generate enough hazard samples from the data sets we have to be able to train models uh to be able to detect critical components railway track damage loose bolts vegetation and how can we do all of that in a representative manner and then just some overall view over here so glo scaling safety for global rail networks so artificial neural networks in particular convolutional neural networks do have a real use case they do have enough within them and there is a lot of supporting frameworks for us now to be able to create applications that are actually useful for the railway industry um and for that what we need to start moving towards is better collaboration with academia industry and at a government level so we can make sure that the standards are in place and we’re getting the fruits for convolution neural networks right so I think I’ve just about uh I’ve gone over one minute I think so I’m sorry for that um so yeah let’s build a safer railway future for everyone thank you okay great thanks for the wonderful presentation very much into AI which is believe it’s a very new field and all over the world people have been exploring more on artificial intelligence now we also have been got some kind of uh computations and uh to participate more on this but then AI in something new as well so thanks for the sharing uh by the way participants have any question to Dr musin well they got one I think in the chat box can I read for you okay salam doctor ai computer vision enhancers safety by detecting obstacles monitoring track etc since AI may require internet connectivity of realities how can we manage or reduce cyber security risk and railway system as far as salam uh okay yeah so we can so that is one of the key areas we’re looking at and now that’s cross domain so our focus is on building the actual architecture which can do the detection so we’re looking at convolution neural networks to be able to say okay where is the damage and how do we detect the damage we then have a layer on top of that which is to be able to materialize it and get it into the production for that is the cyber security side uh you know how do we make sure that the assets what we’re detecting you know stays within the walls of where we want it and is not taken by a third party so there is still work which is being done in terms of getting the cyber security field to work more closely within the creation mechanism of the CNN network to say okay once we’re getting these detections how do we infer them how do we mask them and hash them in a way that we’re able to make sure our whole system is failp proof if you like uh for it respective application but um I think your question there really is how do we get a system which is failp proof and can be deployed on a bigger picture um and I think I agree in the sense that there needs to be more work with uh on a cross domain basis but like I said our focus at the moment is to be able to show that the computer vision part of it is actually working can actually do the detections and can be the key part for us to automate the process then the rest of it is okay how do we put in the you know the rest of the components to make a a shiny product that can be given and can be utilized but for the for the core technology to do the detections computer vision is showing that it can do that okay are you okay with that so shall we go for any other question it’s interesting topic sorry uh partishu I have a question for doctor can go for it yeah okay okay asalam alaikum doctor i’m uh Joe from Indonesia uh I’m very interesting with the AI of the uh for the railway uh we have a in Indonesia we have a tracheometric uh measurement car vehicle car and uh we have uh a million of uh when when uh when the track geometric car u measure inspect uh sorry uh wrong camera okay uh okay uh uh yeah uh and then uh It is a thousand of picture of rail uh but uh I think the can uh uh reduce the time of uh inspection when the the rail is broken or the rail is wearing the warrant so uh uh what uh what kind of computer we have to uh I have to uh invest for this uh uh the the process of the uh the this million picture thank you doctoralamikum um yeah so in terms of I think your question more is on the investment of the hardware now the investment for what you need for hardware really depends on the application um believe it or not a lot of a lot of our work we’ve just been able to do freely based on cloud GPU so Google cloud platform uh utilizing Google collab we’ve been able to do all of our training using free GPU access uh that can also be done through the likes of Amazon Web Services they offer you free GPU access uh Microsoft Azure so a lot of these big cloud platforms understand the benefits now of computer vision AI in general and they’re providing access to GPUs now they’re providing it from the perspective that will give some limited free tier access and then based on that once the GPU runs out uh the developer will say okay I need to subscribe for it i need to pay for it and get more access well there are there are smart ways in the sense of how you utilize your data set how do you prepare your data set you know how smart are you with developing the initial training data set to be able to do all of the process uh in a limited manner without running out of the GPUs that have been allocated to you but with that said uh you know if you do have the investment there and you’re able to make that there’s various different GPUs that can be bought from the market rtx GPU has its own range uh Nvidia gives you a lot of GPU uh choice that can be utilized uh you know for getting your own hardware but yeah feel free to get in touch with me and I’ll be more than happy to discuss that in more detail uh okay okay so uh uh what what kind of software can uh make can uh AI uh uh utilize the uh utilize the uh this mechanism of process it is all it is MATLAB or something I okay so MATLAB is usually used from the engineering domain uh the more optimal way to do it is a platform known as ultralytics so I’ve mentioned that on my slides and all ultralytics there’s another one called Roboflow so these are two platforms which really make the development process quite straightforward so you can upload you know data set do all of the augmentation transformation model training model deployment all of it in a single platform so what you had to do before is use one platform to do the model training another one for augmentation another one for something else another one for deployment these platforms bring it all into one so you could go through the whole process you know just utilizing a single framework okay okay okay okay okay uh thank you doctor for you all right thank you but I beg you it’s already uh 5:45 p.m malaysia time going to be like 6:00 so I believe we got uh this is our first session i’m sure we’re going to have more uh sessions in the future so with that uh thanks again for all the presenters speakers as well as participants for your commitment and you know for giving wonderful participation throughout the session so uh the attendance link have been shared in the chat box kindly fill it up for you to get your e certificate and also uh to end this session we will have a group session photo so kindly please on your uh cam so that the admin will you can give a backshot of yourself so the admin will take a photo as a remembrance and uh in the count of I think will do that so once already please on your camp maybe I give a count one two three one Two three have you got it three yeah all right uh can you hear me yeah you can uh I think uh Dr use need to uh stop the sharing the slide sharing oh sorry right done yeah that’s fine all right can we do it again yeah okay ready everyone um on a count of three one two three smile all right uh one more okay one two three smile okay and another one ready one two three okay all right okay then I think with that we end this international lecture session one with ETHM collaboration with field and ETHM field so uh we will end this session for today so I believe with this we will have more of such kind of session in order to further our collaboration between UTH and of course on the screen okay thank you very much thanks again have a good day take care bye thanks everyone thank you thank you bro thank you