SwinV2 Transformer for Compressed Image Super-Resolution and Restoration

[mv-lab]

trafficstars

Hello,

we love and admire the work your doing :) Recently we tested SwinV2 capabilities for image restoration, compression artifacts removal and super-resolution. The properties it brings are amazing! we can achieve similar results to SwinIR yet with half training time.

Here we leave our paper and results, hopefully it can help to give more visibility to the swin-family.

Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration @ ECCV 2022 AIM Workshop

Github 🚀 https://github.com/mv-lab/swin2sr | code, demos and more insights about the architecture.

Best, Marcos V. Conde

[ancientmooner]

Thanks for your sharing! I suggest that you may also try the Swin V2 models with SimMIM pretraining. In our experience, SimMIM pre-training should be more friendly for low-level vision tasks than that using supervised pre-training.

https://github.com/microsoft/Swin-Transformer/blob/main/MODELHUB.md#simmim-pretrained-swin-v2-models