This is the code our team provided for the Hi Paris 2022 Hackathon computer vision. We won the first place of the Capgemini prize an ranked 3rd in the Innovation category.
The task was to detect and classify cars from a modified version of the Stanford Cars Dataset. We chose to divide the task in two, using a pretrained Microsoft Swin Vision Transformer for detection, and fine tuning a Facebook Research Resnext CNN for classification.
Please run the installation Notebook to have all the dependencies, especially mmcv-full and the pth model.