iVPF: Numerical Invertible Volume Preserving Flow for Efficient Lossless Compression

Shifeng Zhang, Chen Zhang, Ning Kang, Zhenguo Li. iVPF: Numerical Invertible Volume Preserving Flow for Efficient Lossless Compression. In CVPR 2021.

The MindSpore implementation of iVPF.

Dependencies

Python 3.7.5 (Highly recommend version 3.7.5, or unexpected error of Mindspore installation may occur)

Mindspore GPU 1.5 (GPU CUDA 10.1/11.1)

The installation of Mindspore can be found here. The Mindspore version MUST be the newest, or unexperted error on running the code may occur.

Usage

Download model weights here, and put it in folder model_weights.

Inference CIFAR-10 with iVPF

python eval_coding_cifar10.py --data_dir [CIFAR10_DATA_DIR] --no_code

Encoding and Decoding CIFAR-10 with iVPF

python eval_coding_cifar10.py --data_dir [CIFAR10_DATA_DIR]

Reference

@inproceedings{zhang2021ivpf, title={iVPF: Numerical Invertible Volume Preserving Flow for Efficient Lossless Compression}, author={Zhang, Shifeng and Zhang, Chen and Kang, Ning and Li, Zhenguo}, booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition}, pages={620--629}, year={2021} }