Detection and Classification of Vehicles based on Deep Semantic Features of Faster R-CNN
Saleet Ul Hassan, Muhammad Zeeshan Khan, Hafiz Umer Draz, Muhammad Usman Ghani Khan · 2019
In the current era, automation of vehicle detection and classification is considered as one of the challenging tasks which leads to the safety of roads and Transportation system. With the development of Computer vision and Image processing techniques, many researchers overcome the obstacles and achieved the aim. However, they lack accuracy and to handle this issue deep learning technology received a lot of attention from researchers. Deep learning-based Convolution Neural Network Specifically Faster Region (CNN) has been very successful in object detection and Image classification. In the proposed system we have used the latest techniques of deep learning to achieve state of the art results. Proposed system uses Faster R-CNN for both detection and classification purposes, it not only saves the time but also has superiority over other machine learning techniques in sense of accuracy and robustness. Firstly, Region Proposal Network detects the vehicle then CNN classifies the vehicle by its type. We have achieved 97.3% accuracy on dataset generated by ourselves. Achieving such an accuracy on locally generated dataset shows the novelty.