conda env create -f arib_bps.yml

cd src/utils/coder
g++ -O3 -Wall -shared -std=c++11 -fPIC `python -m pybind11 --includes` python_interface.cpp -o mixcoder.so

python compress.py (--encode | --decode) --input [path to input file] --output [path to output file] --config [path to config file] --model [path to model folder]

python test.py --dataset [cifar10|imagenet32|imagenet64|imagenet64_small] --dataset_type [cifar10|imagefolder|filedataset] --data_dir [path to dataset folder] --model [path to model folder] --mode [inference|single|dataset|speed] <--batchsize> [batch size for inference, or size of dataset for dataset compression setting] <--log_path> [logger path for dataset/single compression setting]

python train.py --num_gpus [number of gpus] --mode [sig|ins|finetune] --dataset_type [cifar10|imagefolder|filedataset] --data_dir [path to dataset folder] --valid_num [size of validation set] --config [path to config file] --dropout [probability of dropout] --save_dir [path to save log and weights] --lr [learning rate] --batch_size [batch size for each gpu] --num_iters [number of iterations] --log_interval [interval of logging] --valid_interval [interval of validation] --decay_rate [decay rate of learning rate] --decay_interval [interval of decaying learning rate] <--resume> [path to pretrained checkpoints] <--qp_path> [relative path to qplist, only used for finetune mode] <--master_port> [master port for ddp]

python generate_finetune_dataset.py --dataset_type [cifar10|filedataset] --data_dir [path to dataset folder] --model [path to model folder] --config [path to config file] <--data_dst> [path to folder of qp list and imgs, only required for cifar10 dataset]  <--qp_path> [relative path to qplist]

@InProceedings{zhang2024learned,
    author    = {Zhang, Zhe and Wang, Huairui and Chen, Zhenzhong and Liu, Shan},
    title     = {Learned Lossless Image Compression based on Bit Plane Slicing},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2024},
    pages     = {27579-27588}
}