Follow the instructions in w-plus-adapter to train the generator and e4e encoder. Then, obtain the embeddings of the training images.
Calcultae the class embeddings of the seen categories:
python calculate_class_embeddings.py
Modify the paths in train_sage.sh to your own path.
sh train_sage.sh
To inference on a single image:
python inference_sage.py --input_path <IMAGE PATH>
This project is built based on the excellent w-plus-adapter.