Home / Tools / EDVR

EDVR

Challenge-winning AI video restoration models

Visit EDVR
EDVR preview
github.com Free

Upvotes

0

Pricing

Free

Category

Video Enhancers

In the database since

Oct 2026

Overview

About EDVR

EDVR is a free research project for restoring and enhancing video with artificial intelligence. It was the winning solution in the NTIRE 2019 challenges on video restoration. The method uses enhanced deformable convolutions to align and sharpen frames. It is built for researchers and developers with Python and PyTorch experience. The code has been merged into the BasicSR project. Deformable convolutions are flexible filters that adapt their shape to match motion in the video. Check the official GitHub page for current details and the recommended setup.

Key Features

1. Challenge-Winning Method

Won the NTIRE 2019 video restoration and enhancement challenges. The approach beat competing methods on difficult test footage.

2. Frame Alignment

Aligns neighboring frames before combining them for restoration. Good alignment is key to removing blur and noise cleanly.

3. Merged into BasicSR

The code now lives in the BasicSR super-resolution project. New users should start with BasicSR, which the authors actively maintain.

4. Model Zoo

BasicSR offers a model zoo with pretrained weights for many tasks. Download a trained model instead of training from scratch.

5. Colab Support

A Google Colab notebook is linked for trying the models. Test restoration in the browser without configuring a local environment.

Frequently Asked Questions (FAQ)

What is EDVR?

A video restoration method that won the NTIRE 2019 challenges. It aligns video frames and enhances them using deformable convolutional networks.

Where is the EDVR code now?

It was merged into BasicSR, a larger image and video restoration project. The old repo is kept as a mirror; use BasicSR for new work.

What do I need to run it?

Python, PyTorch, and ideally a GPU. The Colab notebook is the fastest way to try it without local setup.

What tasks does it handle?

Video super-resolution, deblurring, and denoising. BasicSR's model zoo covers these plus single-image restoration tasks.

Can beginners use EDVR?

It is research code for developers. Beginners should start with the Colab demo and the BasicSR documentation.

Conclusion

EDVR is a proven research method for AI video restoration, now continued inside BasicSR. It suits developers and researchers working on enhancement pipelines. Pretrained models and Colab demos make it approachable. Check the official GitHub page today for the BasicSR project and setup guides. Its challenge win shows the method holds up on tough real footage.

Newsletter

Tools like this change fast

New Video Enhancers tools launch every week — the newsletter keeps you ahead of what's worth trying.

Keep exploring

Similar AI tools

Ranked by the job they do, their capabilities, and how closely they fit this workflow.