Recursive Generalization Transformer for Image Super-Resolution
Authors: Zheng Chen, Yulun Zhang, Jinjin Gu, Linghe Kong, Xiaokang Yang
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate that our RGT outperforms recent state-of-the-art methods quantitatively and qualitatively. Code and pre-trained models are available at https://github.com/zhengchen1999/RGT. |
| Researcher Affiliation | Collaboration | Zheng Chen1, Yulun Zhang1 , Jinjin Gu2,3, Linghe Kong1 , Xiaokang Yang1 1Shanghai Jiao Tong University, 2Shanghai AI Laboratory, 3The University of Sydney |
| Pseudocode | No | The paper describes its methods but does not include any figure, block, or section explicitly labeled 'Pseudocode' or 'Algorithm'. |
| Open Source Code | Yes | Code and pre-trained models are available at https://github.com/zhengchen1999/RGT. |
| Open Datasets | Yes | Following recent works (Zhang et al., 2021; Magid et al., 2021; Liang et al., 2021), we choose DIV2K (Timofte et al., 2017) and Flickr2K (Lim et al., 2017) as the training data. For testing, we use five standard benchmark datasets: Set5 (Bevilacqua et al., 2012), Set14 (Zeyde et al., 2010), B100 (Martin et al., 2001), Urban100 (Huang et al., 2015), and Manga109 (Matsui et al., 2017). |
| Dataset Splits | No | The paper does not explicitly provide details about training/validation/test dataset splits, only mentioning training and testing datasets. |
| Hardware Specification | Yes | We use PyTorch (Paszke et al., 2019) to implement our models with 4 Nvidia A100 GPUs. |
| Software Dependencies | No | The paper mentions 'PyTorch (Paszke et al., 2019)' but does not specify its version number or versions for other software dependencies. |
| Experiment Setup | Yes | We train our models with batch size 32, where each input image is randomly cropped to 64x64 size, and the total training iterations are 500K. Training patches are augmented using random horizontal flips and rotations with 90°, 180°, and 270°. To keep fair comparisons, we adopt Adam optimizer (Kingma & Ba, 2015) with β1=0.9 and β2=0.99 to minimize the L1 loss function following previous works (Zhang et al., 2018a; Dai et al., 2019; Liang et al., 2021). The initial learning rate is set as 2x10^-4 and reduced by half at the milestone [250K,400K,450K,475K]. |