Awesome

CPASSRnet-Tensorflow

Simple Tensorflow implementation of "Cross Parallax Attention Network for Stereo Super-resolution"

Requirements

• Tensorflow 1.9
• Python 3.6
• numpy
• scipy
• scikit-image
• pillow
• pandas
• vgg19 pretrained model, download URL : http://www.vlfeat.org/matconvnet/models/imagenet-vgg-verydeep-19.mat

Usage

├── datasets
   └── YOUR_DATASET_NAME
       ├── train_patches_96_288 (patch size 96x288)
           ├── xxx_L.jpg (format doesn't matter, but 'name_L','name_R' must be corresponding. e.g. patch_1_L.png and patch_1_R.png)
           ├── xxx_R.png
           ├── yyy_L.jpg
           └── ...
       ├── test
           ├── aaa_L.jpg 
           ├── aaa_R.png 
           ├── bbb_L.png
           └── ...
├── pretrained_vgg
   └── imagenet-vgg-verydeep-19.mat

Train

• python main.py --phase=train --dataset=middlebury --batch_size=6 --visible_devices=0

Test

• python main.py --phase=test --dataset=middlebury --visible_devices=0

Related works

• PASSRnet-Pytorch
• StereoSR-Tensorflow

Reference & Thanks

• MUNIT-Tensorflow

Author

Canqiang Chen