Car Damage Classification Based on Deep Learning

Abstract

Nowadays, the rise of automobile industries is directly related to the increasing number of car accidents. So, organizations like insurance companies and others are facing a lot of claims leakage and problems of delay in processing it. To reduce the load, artificial intelligence based on machine learning and deep learning algorithms can help to solve these kinds of problems. In this project, we are going to apply deep learning-based algorithms like VGG16 and VGG19 to classify damage and also used dataset which is manually augmented. The algorithms used helps in detecting damaged parts of a car and assess its location and severity. Initially, we discover the effect of domain-specific pre-trained CNN models, which are trained on the ImageNet dataset and followed by fine-tuning along with CAM (Class Active Mapping). Then we apply transfer learning in trained VGG models and used some techniques to increase the accuracy of our system. Based on the performance of VGG models, we will come to know which VGG model is better for our system.  After analyzing and implementing our models, we came to know that the results of using which technique can work better and will give better accuracy. This project will reduce the human efforts and give prompt results.