Running on your own dataset:

1. We used pytorch/pytorch:1.1.0-cuda10.0-cudnn7.5-devel docker. Directory with input images is mounted as /dataset, results will be saved in /output, directory with code is mounted as /TGA. You should set --shm-size 8G parameter.
2. Enter directory /TGA. See TGA.sh script for installation commands and running example.

Video Super-resolution with Temporal Group Attention (TGA)

The official source code (partially cleaned) for the [Video Super-resolution with Temporal Group Attention] which is accepted by [CVPR-2020].

Train

We utilize 8 Nvidia Tesla V100 GPUs for training.

python main.py

Test

cd code
unzip TGA-without-align-dla.zip

We utilize 1 P100 GPU for testing. Test the trained model with best performance by

python test.py