Cross-Scale Internal Graph Neural Network for Image Super-Resolution

Authors: Shangchen Zhou, Jiawei Zhang, Wangmeng Zuo, Chen Change Loy

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate the effectiveness of IGNN against the state-of-the-art SISR methods including existing non-local networks on standard benchmarks.
Researcher Affiliation Collaboration Shangchen Zhou1 Jiawei Zhang2 Wangmeng Zuo3 Chen Change Loy1 1Nanyang Technological University 2Sense Time Research 3Harbin Institute of Technology {s200094,ccloy}@ntu.edu.sg zhangjiawei@sensetime.com wmzuo@hit.edu.cn
Pseudocode No The paper does not contain any pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes https://github.com/sczhou/IGNN
Open Datasets Yes Following [23, 12, 45, 43, 5], we use 800 high-quality (2K resolution) images from DIV2K dataset [34] as training set.
Dataset Splits No The paper does not explicitly provide training/validation/test dataset splits. It mentions using DIV2K for training and evaluating on standard benchmarks, but no specific validation split information.
Hardware Specification Yes The IGNN is implemented on the Py Torch framework on an NVIDIA Tesla V100 GPU.
Software Dependencies No The paper mentions 'Py Torch framework' but does not specify a version number or other software dependencies with their versions.
Experiment Setup Yes We set the minibatch size to 4 and train our model using ADAM [18] optimizer with the settings of β1 = 0.9, β2 = 0.999, ϵ = 10 8. The initial learning rate is set as 10 4 and then decreases to half for every 2 105 iterations. Training is terminated after 8 105 iterations. The network is trained by using ℓ1 norm loss.