Create the virtual environment according to env.yaml and running app.py.
Both Dockerfile and start.sh are about Docker.
Image dataset is spu dataset. All_images is not uploaded which contains all images.
Train.py is a training file, and you need to change the path. Resnet34 is a trained model with an accuracy rate of over 93% on the test set.
Resnet_ym is a rewritten model that outputs the 512-dimensional output of the last linear layer to build the vector library.
resnet34v512.bin can be gotten from https://drive.google.com/file/d/1ddcMjpRd5tGta4pxCs5HrWk72ACiUvEl/view?usp=share_link