Deep Learning Based Car Damage Classification and Detection

EasyChair Preprint 3008

Abstract

In this paper, we designed and implemented a car damage classification/detection pipeline, which can be used by insurance companies to automate the process of vehicle insurance claims. The recent advances in computer vision largely due to the adoption of fast, scalable and end to end trainable Convolution Neural Networks(CNN’s) makes it technically feasible to recognize vehicle damages using deep convolution networks.We manually collected and annotated images from various online sources using web crawler containing different types of vehicle damages. Due to the relatively small size of our dataset, we used models per-trained on a large and diverse dataset to avoid over-fitting and learn more general features. Using CNN models pretrained on ImageNet dataset and applying various techniques to improve the performance of our system, we were able to achieve accuracy of 96.39%, significantly better than results achieved in the past on a similar test-set.Furthermore to detect the region of damage we used state-of-the-art YOLO object detector and achieving a maximum map score of 77.78 % on the held-out test set, demonstrating that the model was able to successfully recognize different vehicle damages. In addition to this , we also propose a pipeline for a more robust identification of the damage in vehicles by combining the tasks of classification and detection. Overall these results pave the way for further research in this problem domain and we believe collection of a more diverse dataset would be sufficient to implement an automated vehicle damage identification system in the near future.

Keyphrases: Car damage classification/detection, Convolutional Neural Network, MAP score, Pre-trained CNN models, Transfer Learning, YOLO object detector, computer vision, damage detection

@booklet{EasyChair:3008,
  author    = {Hashmat Shadab Malik and Mahavir Dwivedi and S. N. Omakar and Satya Ranjan Samal and Aditya Rathi and Edgar Bosco Monis and Bharat Khanna and Ayush Tiwari},
  title     = {Deep Learning Based Car Damage Classification and Detection},
  howpublished = {EasyChair Preprint 3008},
  year      = {EasyChair, 2020}}