cydiachen / SCET

Self-Calibrated Efficient Transformer for Lightweight Super-Resolution (Official-Deployment)

Self-Calibrated Efficient Transformer for Lightweight Super-Resolution (official-Deployment)

Introduction

We have implemented our SCET method through the mmediting and mim algorithm framework. Next, we will describe the main processes of training and testing.

• Paper The SCET has been accepted by CVPRW2022, you can read the paper here.
• Model

In this repo, we will provide detailed code for deploy this image super-resolution model in ONNXRuntime, TensorRT and other platform.

Training Details

You can refer to our paper version repo SCET for reimplementing this paper and train your own image super-resolution network.

This repo mainly provides the detailed implementation for various kind of devices.