Vehicle Identification Using RESNET-50: CNN approach

Vehicle This paper proposes a method for vehicle brand identification using a Convolution Neural Network (CNN) with improved Accuracy. The proposed method utilizes static image datasets and live CCTV surveillance data for training and real-time vehicle detection and recognition. This system can detect moving vehicles and recognize their brands, count them, and verify permits. Leveraging the advancements in deep learning, the proposed model builds upon classical CNNs with specific enhancements. This improved model is trained on a benchmark vehicle detection dataset and demonstrates superior performance in identifying vehicles on the road. The primary objective of this study is to develop a cost-effective and efficient solution for identifying vehicle brands across various applications. A deep learning-based classification technique is employed to achieve this goal. The model utilizes the popular ResNet-50 architecture combined with transfer learning techniques. Initial training yielded an accuracy of 34%. However, with further training, the model improved significantly, reaching an accuracy of 95.31%. A training-to-testing data split ratio of 43:7 was used.