ReMoNet: Recurrent Multi-Output Network for Efficient Video Denoising

Authors: Liuyu Xiang, Jundong Zhou, Jirui Liu, Zerun Wang, Haidong Huang, Jie Hu, Jungong Han, Yuchen Guo, Guiguang Ding2786-2794

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments on NVIDIA GPU and Qualcomm Snapdragon 888 mobile platform with Gaussian noise and simulated Image Signal-Processor (ISP) noise. The experimental results show that Re Mo Net is both effective and efficient on video denoising.
Researcher Affiliation Collaboration 1 Beijing National Research Center for Information Science and Technology (BNRist) 2 School of Software, Tsinghua University, Beijing, China 3 OPPO Inc, Guangdong, China. 4 Computer Science Department, Aberystwyth University, SY23 3FL, UK
Pseudocode No The paper does not contain any 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 code repository for the described methodology.
Open Datasets Yes We use two benchmark datasets for evaluation: Set8 and DAVIS-test. Set8 consists of 4 sequences captured by GOPRO camera and 4 sequences from the Derf s Test Media collection, with a resolution of 960 × 540. The DAVIS-test contains 30 sequences of resolution 854 × 480. We use DAVIS-train for training.
Dataset Splits Yes We use two benchmark datasets for evaluation: Set8 and DAVIS-test. [...] We use DAVIS-train for training.
Hardware Specification Yes We conduct extensive experiments on NVIDIA GPU and Qualcomm Snapdragon 888 mobile platform with Gaussian noise and simulated Image Signal-Processor (ISP) noise. The experiments are conducted on the Qualcomm Snapdragon 888 mobile platform, where its GPU is used for inference.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, CUDA versions).
Experiment Setup Yes In practice, we choose the input number of frames 2T + 1 = 5 and temporal fusion size 2K + 1 = 3, the RTF hidden dimension L = 32.