Functional Neural Networks for Parametric Image Restoration Problems

Authors: Fangzhou Luo, Xiaolin Wu, Yanhui Guo

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In Section 4, extensive experiments are conducted to evaluate Func Nets on three parametric image restoration tasks.
Researcher Affiliation Academia Fangzhou Luo Mc Master University luof1@mcmaster.ca Xiaolin Wu Mc Master University xwu@mcmaster.ca Yanhui Guo Mc Master University guoy143@mcmaster.ca
Pseudocode No The paper describes the methods in text but does not include structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide an explicit statement about releasing source code or a link to a repository.
Open Datasets Yes Following [67, 23, 60], we use the DIV2K dataset [49] for training.
Dataset Splits Yes There are 1000 high-quality images in the DIV2K dataset, 800 images for training, 100 images for validation and 100 images for testing.
Hardware Specification No All experiments run in parallel on 4 GPUs.
Software Dependencies No No specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow, specific library versions) are provided.
Experiment Setup Yes All our three Func Net models are trained by ADAM optimizor with β1 = 0.9, β2 = 0.999, and ϵ = 10 8. The initial learning rate is set to 10 4 and then decreases by half for every 2 105 iterations of back-propagation. In the super-resolution problem, we randomly extract 32 LR RGB patches with the size of 40 40 as a batch input. In the image denoising problem, we randomly extract 32 RGB patches with the size of 96 96 as a batch input. In the JPEG deblocking problem, we randomly extract 32 gray patches with the size of 96 96 as a batch input