Insurance companies spend valuable time and resources identifying vehicle damage to properly asses claims. Using platform.ai, we trained a neural network to classify car damage location with around 80% accuracy. We used a dataset of web-scraped images, including pictures of front, side, and rear damage to a car. Using platform.ai’s projection tool, we selected clusters of similar images, applied some labels, and started training our model. By repeating this process a few times, we were able to train a deep neural network without having to code anything!
