• CAE-ADMM
• • Our Contributions
  • File Structure
  • Environment
  • Command
  • Train
  • test
  • Comparison
  • • Reconstructed image
    • Quality measurements
  • Citation
  • Contributors

CAE-ADMM

CAE-ADMM: Implicit Bitrate Optimization via ADMM-based Pruning in Compressive Autoencoders

key words: Image Compression, Compressive Autoencoders, Adam

The model utilizes the improved Compressive Autoencoders to compress images. Also, the paper proposes an Adam-based method to optimize the process. The paper is published in 2019, and readers can read the original paper via the link.

Our Contributions

1. Translate from Pytorch version to mindspore
2. Successfully run forward, backward and parameter update
3. Test the mindspore version, and compare it with the original version

File Structure

The translated model file is in ./demo/model_baseline.py.

demo
├── ckpt
│   └── latest.ckpt      # test model file
├── model_baseline.py    # reproduce model
├── huffmancoding.py     # huffman coding
├── test.py              # test the performance
└── train.py             # train

Environment

• mindspore-dev==2.0.0dev20230116
  • It is recommended to install through the link.
• cuda 11.1
• python 3.7

Command

Please intall compressai via

pip install -e .

under dictionary CompressAI_MindSpore.

Train

Train the model via:

python train.py -d your/own/dataset/address

test

We code and decode via:

python test.py -d your/own/dataset/address

Actually, it is very fast although it is related to disk I/O. Therefore, we will not provide the fast version of testing.

Comparison

Reconstructed image

Mindspore version: PSNR (dB): 28.703 | MS-SSIM: 0.9618 | SSIM: 0.8547 | bpp: 0.792

Quality measurements

	MS-SSIM	SSIM	bpp
Official version	0.9691	0.8747	0.5407
Mindspore version	0.9711	0.8674	0.800
Mindspore version	0.9791	0.9304	1.482
Official version	0.9849	0.9381	1.012

Citation

@article{zhao2019cae,
  title={CAE-ADMM: Implicit Bitrate Optimization via ADMM-based Pruning in Compressive Autoencoders},
  author={Zhao, Haimeng and Liao, Peiyuan},
  journal={arXiv preprint arXiv:1901.07196},
  year={2019}
}

Contributors

Name:

Zhuozhen Yu

Hua Ye