ResDiff: Combining CNN and Diffusion Model for Image Super-resolution

Authors: Shuyao Shang, Zhengyang Shan, Guangxing Liu, LunQian Wang, XingHua Wang, Zekai Zhang, Jinglin Zhang

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The extensive experiments on multiple benchmark datasets demonstrate that Res Diff outperforms previous diffusion-based methods in terms of shorter model convergence time, superior generation quality, and more diverse samples.
Researcher Affiliation Academia Shuyao Shang 1, Zhengyang Shan 1, Guangxing Liu 1, Lun Qian Wang 2, Xing Hua Wang 2, Zekai Zhang 3, Jinglin Zhang 1 1 Shandong University 2 Linyi University 3 Qilu University of Technology
Pseudocode Yes Algorithm 1: Res Diff Inference
Open Source Code No The paper does not provide any concrete links or explicit statements about the release of its source code.
Open Datasets Yes Experiments on two face datasets (FFHQ and Celeb A) and two general datasets (Div2k and Urban100) demonstrate that Res Diff not only accelerates the model s convergence speed but also generates more fine-grained images.
Dataset Splits Yes Our Res Diff is trained solely on the provided training data to guarantee a fair comparison.
Hardware Specification No The paper mentions 'Due to equipment limitations' in the conclusion but does not specify the hardware (GPU/CPU models, memory, etc.) used for running the experiments.
Software Dependencies No The paper does not explicitly list specific software dependencies with version numbers (e.g., Python, PyTorch, CUDA versions).
Experiment Setup No The supplementary material contains detailed information about the training process, hyperparameters, and other relevant details.