Identification of Car Make and Model Using Deep Learning and Computer Vision Techniques

Citations Over Time

Abstract

Vehicle model detection is now an integral part of the traffic scene. In the realm of object detection, Deep Learning models have been widely applied. Though there are few research articles in this area, the lack of effective image recognition has been a major challenge to be addressed. Therefore, the present study has used Single Shot Detector (SSD)-YOLOV5 for car recognition and a deep residual network (ResNet) for classification purposes to combat this issue. The proposed model will take the real-time image as an input and using probability estimation, attempts to recognize the car model. We used a combination of the Stanford University car dataset with 196 classes and an Indian car dataset with 97 classes to train the model. The proposed model achieved an accuracy of 86%, which was higher compared to the VGG16 model's accuracy of 71%. The front-end web application was constructed with React JS and the model is deployed using the Flask micro-framework.

Related Papers

• → Face recognition classifier based on dimension reduction in deep learning properties(2017)7 cited
• → A Solution to Improve Object Detection for Images Captured by UAV-mounted Camera(2019)3 cited
• → Fusing different levels of deep features by deep stacked neural network for hyperspectral images(2017)1 cited
• → Multiclass Object Recognition Using Object-based and Local Image Features Extractors(2015)