This repository shows a quick demo for how to upscale videos downloaded from YouTube using the implementation of "Investigating Tradeoffs in Real-World Video Super-Resolution, arXiv". Code has been modified from the official repo.
- How Super Resolution Works
- MMEditing. MMEditing is an open source image and video editing toolbox based on PyTorch.
- List of accepted papers in the "New Trends in Image Restoration and Enhancement" workshop (CVPR2021)
- BasicVSR++. An improvement over RealBasicVSR from the same authors.
- Image Super-Resolution via Iterative Refinement. Code. Google Research.
- Two-minute Papers Review on Super-Resolution
- SwinIR: Image Restoration Using Swin Transformer. HuggingFace Spaces
- VRT: A Video Restoration Transformer. Repo. From the same authors of
SwinIR. Achieves SoTA (up to 2.16dB) in video SR (REDS, Vimeo90K, Vid4 and UDM10), video deblurring (GoPro, DVD and REDS), video denoising (DAVIS and Set8) - Waifu2x-Extension-GUI. Photo/Video/GIF enlargement and Video frame interpolation using machine learning. Only runs on Windows.
Two videos in YouTube (short duration, low input quality).
Here are some before and after images that have been processed through RealBasicVSR. Depending on how different the test data is from the trainig data used, results will vary. The VideoLQ-Dataset can be explored and downloaded here.
- SageMaker Studio Lab Account. Free GPU (T4) for up to 4 hours at a time. See video for more info.
- Python 3.7+
- PyTorch >= 1.7.1 and Torchvision >= 0.8.2. Official instructions
- mim and mmcv-full
- youtube-dl
- see
requirements.txt
- Click the following button to open the sample Notebook
- Once opened, click on
Copy to Projectto clone the repo into Studio Lab. Because we have included anenvironment.ymlfile, Studio Lab will automatically build a Conda environment with all required dependencies. It will be named asmachinelearnear-RealBasicVSR-youtubeand will be selected by default when you open the sample Notebook.
@article{chan2022investigating,
author = {Chan, Kelvin C.K. and Zhou, Shangchen and Xu, Xiangyu and Loy, Chen Change},
title = {Investigating Tradeoffs in Real-World Video Super-Resolution},
journal = {IEEE Conference on Computer Vision and Pattern Recognition},
year = {2022}
}