Vehicle damage inspection through deep learning

This was a Minimum Viable Product built using Instance Segmentation. Frontend was built on React Web framework. Backend developed using FastAPI and database on PostgreSQL. We used Tensorflow Keras for deep learning and used Mask RCNN for Instance Segmentation to mask the vehicle damage output. Before passing the image to the main module our damage detector engine would do some quality control to ensure that vehicle image is authentic and used some additional Third Party API's to fetch the vehicle details from their database which would be helpful in estimating the damage claim cost.