This is the initial version of the Real-RSCM dataset and CMA-WSA Framework.

conda create -n CMA-WSA python=3.11
conda activate CMA-WSA

install pytorch

# e.g. CUDA 11.8
# with conda
conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
# with pip
pip install torch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 --index-url https://download.pytorch.org/whl/cu118

pip install -r requirements.txt

• Datasets V1.0 is released at Baidu Cloud and Google Drive (2024.12.25)
• Dataset Directory: datasets/
• Dataset Subdirectory: datasets/JsonFiles/, datasets/image/, datasets/source/, datasets/target/, datasets/background/, datasets/segmentation/

Download clip-b-32 weights from Hugging Face

• Clip Directory: models/clipModels/openai_clip_b_32/

Download U-Net weights from Github

• U-Net Directory: models/imageModels/milesial_UNet/

python main.py

• Modify the experiment settings and hyperparameters in src/config.py

CC BY-NC-SA 4.0

All images and their associated annotations in Global-TQA can be used for academic purposes only, but any commercial use is prohibited.