The project aims to develop a Road Asset Detection system using convolution neural network. We had modified existing tiny YOLO model for the detections.
Our project is an extended work on the blooming Computer Vision on localization of relevant objects. We implement a real-time system capable of accounting for different road assets like traffic signs,potholes,lanes,zebra lines and vehicles . We have used the YOLO model for this purpose.We used this model as they have superior localization and classification in addition to easier modification of layers,via transfer learning.We further seek to implement a fully functional real time system on Indian Roads.
Pascal VOC: 2029 Potholes : 630 LISA: 6618
Pothole Detection :84 Vehicles :88 Traffic Signs:87
This accuracy can be increased further by adding more images to the dataset and training them further.