README.md

Sliding Window Recurrent Network for Efficient Video Super-Resolution

Requirements

• Python: 3.6, 3.7
• Python packages: numpy, imageio and pyyaml
• TensorFlow >= 2.6.0 + CUDA cuDNN
• GPU for training (e.g., Nvidia GeForce RTX 2080)

Overview Figure

Dataset preparation

• Download REDS dataset and modify the data_dir in config.yml.

Training and Validation

Training

To train the model, use the following command:

python run.py --process train --config_path config.yml

After training, the checkpoints will be produced in log_dir.

Validation

To valid the model, use the following command:

python run.py --process test --config_path config.yml

After validation, the output images will be produced in log_dir/output.

Testing

To generate testing outputs, use the following command:

python generate_output.py --model_path model/mobile_rrn.py --model_name MobileRRN --ckpt_path snapshot/ckpt-98 --data_dir /data/dataset/aim22/reds/test/test_sharp_bicubic/X4/ --output_dir results

After testing, the output images will be produced in results.

Convert to tflite

To convert the keras model to tflite, use the following command:

python convert.py --model_path model/mobile_rrn.py --model_name MobileRRN --input_shapes 1,320,180,9:1,320,180,32 --ckpt_path snapshot/ckpt-98 --output_tflite tflite/model.tflite

Citations

If our code helps your research or work, please consider citing our paper.

@article{lian2022sliding,
  title={Sliding Window Recurrent Network for Efficient Video Super-Resolution},
  author={Lian, Wenyi and Lian, Wenjing},
  journal={arXiv preprint arXiv:2208.11608},
  year={2022}
}