Ju Rui-Yang ; Chen Chih-Chia ; Chiang Jen-Shiun ; Lin Yu-Shian ; Chen Wei-Han ; Chien Chun-Tse

Abstract

The Transformer-based method has demonstrated remarkable performance forimage super-resolution in comparison to the method based on the convolutionalneural networks (CNNs). However, using the self-attention mechanism like SwinIR(Image Restoration Using Swin Transformer) to extract feature information fromimages needs a significant amount of computational resources, which limits itsapplication on low computing power platforms. To improve the model featurereuse, this research work proposes the Interval Dense Connection Strategy,which connects different blocks according to the newly designed algorithm. Weapply this strategy to SwinIR and present a new model, which named SwinOIR(Object Image Restoration Using Swin Transformer). For image super-resolution,an ablation study is conducted to demonstrate the positive effect of theInterval Dense Connection Strategy on the model performance. Furthermore, weevaluate our model on various popular benchmark datasets, and compare it withother state-of-the-art (SOTA) lightweight models. For example, SwinOIR obtainsa PSNR of 26.62 dB for x4 upscaling image super-resolution on Urban100 dataset,which is 0.15 dB higher than the SOTA model SwinIR. For real-life application,this work applies the lastest version of You Only Look Once (YOLOv8) model andthe proposed model to perform object detection and real-life imagesuper-resolution on low-quality images. This implementation code is publiclyavailable at https://github.com/Rubbbbbbbbby/SwinOIR.