Zero-Shot Image Restoration Using Denoising Diffusion Null-Space Model

Authors: Yinhuai Wang, Jiwen Yu, Jian Zhang

ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on several IR tasks reveal that DDNM outperforms other state-of-the-art zero-shot IR methods. We also demonstrate that DDNM+ can solve complex real-world applications, e.g., old photo restoration. and 4 EXPERIMENTS Our experiments consist of three parts. Firstly, we evaluate the performance of DDNM on five typical IR tasks and compare it with state-of-the-art zero-shot IR methods. Secondly, we experiment DDNM+ on three typical IR tasks to verify its improvements against DDNM. Thirdly, we show that DDNM and DDNM+ perform well on challenging real-world applications.
Researcher Affiliation Academia 1Peking University Shenzhen Graduate School, 2Peng Cheng Laboratory
Pseudocode Yes Algorithm 1 Sampling of DDNM and Algorithm 2 Sampling of DDNM+
Open Source Code Yes Code is available at https://github.com/wyhuai/DDNM.
Open Datasets Yes We choose Image Net 1K and Celeb A-HQ 1K datasets with image size 256 256 for validation.
Dataset Splits No The paper uses pre-trained denoising networks and mentions using ImageNet 1K and Celeb A-HQ 1K for validation, but it does not specify the train/validation/test splits (percentages or counts) that were used for their specific experiments, nor does it provide citations to such predefined splits in a way that allows direct reproduction of the data partitioning.
Hardware Specification Yes on a single 2080Ti GPU with batch size 1
Software Dependencies No The paper provides 'Pytorch-like codes' in Appendix E but does not specify version numbers for PyTorch or any other software dependencies.
Experiment Setup Yes We use DDIM as the base sampling strategy with η = 0.85, 100 steps, without classifier guidance, for all diffusion-based methods. and For fair comparison, we set T = 250, l = s = 20, r = 3 for DDNM+ while set T = 1000 for DDNM so that the total sampling steps and computational consumptions are roughly equal.