Realtime pothole detection - A Complete Implementation

ahead with let's go ahead and see our demo and implementation let's go ahead and see our demo and implementation the complete implementation is live and it's real time it is fast accurate the classification is done in such a way that everybody can understand it very nicely and the voice alert that we give is also going to be very understandable for the driver to be careful about the great three part rules the grade three part rules are more severe we have taken our algorithms in such a way that it can detect the potholes depth and width and based on which we will classify the portals as severe mild non-severe and based on that the grading and the color system color coding has also been given that way we will show you the demo right now and I am sure you will like it for the scene one data set which has been given by the organizers we have implemented this and you can see that grade 1 and grade 2 will not give you any audio Castle [Music] so for the grade three we alert the drivers with the audio alert saying that be careful when you are driving and this is fully implemented you can see the way the system is behaving in a real-time and responding properly we have muted the alert right now because we would like to explain the system and this feature of audio alert is there for grade 3 and if you want we can move it for grade two as well the entire scene is properly analyzed and the potholes are detected real time and this alert goes to the driver and I hope you like the way the system is working on for the scene one that you have given the scene 2 is also completely implemented and there also this grading is working fine and we are very sure our system will work in detection of patrols at various levels various roads and it's highly versatile our algorithm is implemented in such a way that it can be easily ported for other situations as well you can see the exact way the system is identifying the pothole's real time and thank you very much for the data set and the images it was very clean and easy for us to go ahead with implementing and we would like to thank the organizers for that particular aspect of providing us a very neat data set it is all real time and computationally this is not so intensive and it will not take a lot of time for you to detect the patroller algorithms are smart enough that you can detect things with these okay you could see the way the potholes are identified and they are graded as well the color coding is also given such a way that anybody can understand based on the color green no problem yellow be a little careful red you must be definitely careful So based on the length depth and the width of the pothole we are classifying it and it has come through superb level of trainings that we have given into our system so the system is very well trained and it can adapt to any situation you can go ahead of the scene to implementation right now and I am sure you will like it too great ones again in the green color grade 2 yellow and the grade 3 will come with a red color identification the bounded box with the grade information there comes in front of you and grade 3 is highlighted very strongly and we also have the voice assist for grade three if you want we can provide that for grade 2 as well it's based on your inputs we can work on and we can provide you that kind of support and you could see that it is detecting classifying very clearly and all this at a very good speed we can implement this at the edge also and we can make sure that this gives the best results even at the edge the complete implementation is presented to you for your kind perusal and if you have suggestions we are more than open to implement it and we would be very happy to listen to your feedback and suggestions about this implementation

Channel: Shriram Vasudevan

Views: 2,013

Keywords: Innovative Projects, Deep learning, Pothole Detection, Realtime pothole detection, Deep Learning based Pothole detection, DL, Shriram Vasudevan

Id: Te9Vccc5e8I

Length: 3min 45sec (225 seconds)

Published: Tue Feb 14 2023