DifFSS: Diffusion Model for Few-Shot Semantic Segmentation

This is the implementation of the DifFSS method on BAM (DifFSS-BAM) in paper "DifFSS: Diffusion Model for Few-Shot Semantic Segmentation".

For more information, Please refer to the the paper on [arXiv].

🛠️ Getting Started with DifFSS-BAM

1. Preparation of environment and dataset

Please configure the conda environment and prepare the required datasets and base model weights as described in the readme file of the original BAM model.

2. Preparation of auxiliary support images

Please refer to ControlNet4FSS for the pre-generation of auxiliary supoort.

3. Traing

Once the pre-generated auxiliary support images are ready, you can start training the model by modifying the aux_data_dir field in the config file.