Conditional Diffusion Compresion

A Reporece of the project "Conditional Diffusion Compression" Original project Link: https://github.com/buggyyang/CDC_compression

requirements

conda env create -f environment.yml

Dataset

The dataset is from vimeo-90k dataset. http://toflow.csail.mit.edu/

Due to the server's file number limitation, I converted the images to NumPy format and saved them in an h5 file while extracting the dataset. As a result, all the images are stored in train.h5 and val.h5.

# Download the dataset
wget http://data.csail.mit.edu/tofu/dataset/vimeo_septuplet.zip

# unzip the dataset
# need to change the path variable in the unzip_vimeo.py
python data/pre_process_dataset/unzip_vimeo.py

Training

python train.py

Testing

python test_xparam.py