Tensor FISTA-Net for Real-Time Snapshot Compressive Imaging

Authors: Xiaochen Han, Bo Wu, Zheng Shou, Xiao-Yang Liu, Yimeng Zhang, Linghe Kong10933-10940

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on synthetic datasets show that the proposed Tensor FISTA-Net achieves average PSNR improvement of 1.63 3.89d B over the state-of-the-art algorithms.
Researcher Affiliation Academia Xiaochen Han,1 Bo Wu,2 Zheng Shou,2 Xiao-Yang Liu,2 Yimeng Zhang,2 Linghe Kong1 1Shanghai Jiao Tong University, China 2Columbia University, USA
Pseudocode No The paper describes the iterative steps and modules of Tensor FISTA-Net through mathematical equations and descriptions, but does not present them in a structured pseudocode or algorithm block.
Open Source Code Yes The code is available at https://github.com/Guillermo Han97/SCI-Tensor-FISTA-Net.
Open Datasets Yes We evaluate the proposed Tensor FISTA-Net on three different synthetic datasets: Kobe (Yang et al. 2014), Park (Ma et al. 2019) and Vehicle (Ma et al. 2019).
Dataset Splits No Each testing dataset contains 32 frames of size 256 256 and B = 8, i.e., 4 measurements. We use the same training videos NBA, Central Park Aerial and Vehicle Crashing Tests in (Ma et al. 2019).
Hardware Specification Yes We do all the experiments on a server with an NVIDIA Telsa V100-PCIE GPU (16GB device memory).
Software Dependencies No We use Tensor Flow to implement our algorithm and do all the experiments on a server with an NVIDIA Telsa V100-PCIE GPU (16GB device memory). ... Adam optimizer (Kingma and Ba 2014) is used to minimize the training loss.
Experiment Setup Yes We set the number of phases as 10, learning rate as 0.0001 and running epoch as 500. Adam optimizer (Kingma and Ba 2014) is used to minimize the training loss.