mustaphau/ai-video-super-resolution

GDAVSR

AI Video Super-Resolution

Upscale video resolution quality with advanced AI algorithms

The process of reconstructing high-resolution video from low-resolution video is called super-resolution reconstruction. Recent state-of-the-art super-resolution methods have achieved impressive performance on ideal datasets. However, these methods always fail in real-world video super-resolution, since most of them ignore the blurring and noise that exists in the real world (such as video compression). Together with scientists from the GluonCV team, we developed a deep learning based super-resolution approach named GDAVSR. This approach can bring better visual quality. Based on this method, we can help customers improve the resolution of videos at a low cost.

Architecture diagram

Usage

Prerequisites

• An AWS account
• Python installed, such as 3.6
• Install dependencies of app
• Build lambda layer

pip install -r requirements.txt
cd layers
chmod +x build_layer.sh
./build_layer.sh
cd ..

CDK Synth & CDK Deploy

cdk synth -c inferentia=true
cdk deploy --parameters MaxvCpus=<max cpu>

MaxvCpus:default: 16

How to use

1. Upload low-resolution video to S3 bucket('superresolutions*')
2. call Lambda func SuperResolutionSplitVideo

{
"key": "video_file",
"scale": "4",
"task": "inference"
}

- key: filename in s3 bucket
- scale['2', '4']: Scale for super resolution. Default: 2
- task[inference, debug]: When it is set to debug, the intermediate file will be uploaded to S3 for debugging. Default: inference
- env[onDemand, spot]: which the type of machine to be used. Default: spot

1. Wait for Batch processing..
2. Download high-resolution video from S3 bucket('superresolutions*')