Using classical and neural image embeddings and finetuned end-to-end networks to achieve top-tier performance on a vehicle type classification task. Containerized and deployed model as a web app
If we end up doing this, use cvat.ai to label the four classes
Upload unblurred images (roughly 8k of them), divide up the work checking in spare time. it should go pretty quickly because cvat will provide an efficient interface
If we end up doing this, use cvat.ai to label the four classes
Upload unblurred images (roughly 8k of them), divide up the work checking in spare time. it should go pretty quickly because cvat will provide an efficient interface