Research on the Recognition of Car Models Based on Deep-learning Networks

Shuai Zhao, Xiang-Lei Zhu, Ying-Bo Li, Lu Zhang


Based on deep-learning and artificial intelligence technology, car models recognition is very practical and could be widely used in automobile inspection, vehicle monitoring and second-hand car trading. Car model recognition is more difficult than vehicle recognition, since it not only to identify the outline of the car, but also need to extract the features which can indicate different car brands and models in depth. In order to achieve this purpose, the algorithm and model accuracy poses a higher requirement. Based on the depth learning network algorithm, the AlexNet convolutional neural network model and the VGG16 convolutional neural network model are constructed based on the Caffe frame, and 6000 car pictures of different brands and models are used as training dataset and 1000 pictures as verification dataset. GPU has also been applied for network training. Finally, experiments show that the accuracy of VGG16 network is 94.7%, which has a great advantage in the problem solving of car models recognition.


Full Text:



  • There are currently no refbacks.