/deepgtv

Unrolling of Deep Graph Total Variation for Image Denoising

Primary LanguagePython

Paper: Unrolling of Deep Graph Total Variation for Image Denoising

GitHub: huyvd7/deepgtv

Fig. Trained and tested on different noise distributions.

Dataset descriptions

See Dataset

Train DGTV

python train_dgtv.py --batch 32 --lr 1e-4 --umin 0.001 --umax 1000 -m model/DGTV.pkl --epoch 50 --train TRAINSET_PATH --width 36

Note:

  • width: square patch size

Test DGTV

python test_dgtv.py -w 720 -m MODEL_NAME --stride 9 --multi 500 -p TESTSET_PATH

Note:

  • stride: # of pixels moved between two patches
  • multi: # of patches to be processed simultaneously

TODO

  • Denoise a given single image