Check this repo for details.
This is the implementation of ther paper,
Yueyu Hu, Wenhan Yang, Jiaying Liu, Coarse-to-Fine Hyper-Prior Modeling for Learned Image Compression, AAAI Conference on Artificial Intelligence (AAAI), 2020
The currently available code is for evaluation, while it can also be modified for training as the implementation of the network is available.
The code requires the TensorFlow library (v1.13, v1.14 and v1.15 tested). It should be running in the CPU-only mode, for example, by specifying CUDA_VISIBLE_DEVICES= . An example to run the encoder and decoder is provided below.
You may first download the trained weights from Google Drive and place the .pk files under the models folder (that is, to make './models/model0_qp1.pk exist).
python AppEncDec.py -h
python AppEncDec.py compress example.png example.bin --qp 1 --model_type 0
python AppEncDec.py decompress example.bin example_dec.png
Detailed command line options are documented in the help mode of the APP.