Video Super Resolution On Msu Vsr Benchmark

평가 지표

1 - LPIPS

ERQAv1.0

FPS

PSNR

QRCRv1.0

SSIM

Subjective score

평가 결과

이 벤치마크에서 각 모델의 성능 결과

모델 이름	1 - LPIPS	ERQAv1.0	FPS	PSNR	QRCRv1.0	SSIM	Subjective score	Paper Title	Repository
RSDN	0.819	0.667	1.961	25.321	0.619	0.826	5.566	Video Super-Resolution with Recurrent Structure-Detail Network
RRN-5L	0.856	0.617	2.74	23.786	0.549	0.789	5.02	Revisiting Temporal Modeling for Video Super-resolution
Real-ESRnet	0.871	0.598	1.019	27.195	0	0.824	3.697	Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data
HCFlow	0.923	0.713	0.066	26.067	0	0.791	4.262	Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling
VRT	0.929	0.758	2.778	31.669	0.722	0.902	7.628	VRT: A Video Restoration Transformer
Real-ESRGAN	0.895	0.663	1.01	24.441	0	0.774	5.392	Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data
LGFN	0.903	0.74	0.667	31.291	0.629	0.898	6.505	Local-Global Fusion Network for Video Super-Resolution
COMISR	0.879	0.654	1.613	26.708	0.619	0.84	5.637	COMISR: Compression-Informed Video Super-Resolution
RealSR	0.911	0.69	0.352	25.989	0	0.767	5.286	Real-World Super-Resolution via Kernel Estimation and Noise Injection
waifu2x-cunet	0.861	0.6	1.282	27.716	0	0.838	3.308	-	-
waifu2x-anime	0.856	0.57	3.125	27.176	0	0.819	2.739	-	-
TDAN	0.721	0.706	0.493	30.244	0.609	0.883	5.454	TDAN: Temporally Deformable Alignment Network for Video Super-Resolution
DFDnet	0.623	0.339	0.909	24.832	0	0.759	0.277	Blind Face Restoration via Deep Multi-scale Component Dictionaries
TMNet	0.931	0.712	1.136	30.364	0.549	0.885	6	Temporal Modulation Network for Controllable Space-Time Video Super-Resolution
DUF-28L	0.87	0.645	0.418	25.852	0.549	0.83	5.324	Deep Video Super-Resolution Network Using Dynamic Upsampling Filters Without Explicit Motion Compensation	-
SOF-VSR-BI	0.904	0.66	0.571	29.381	0.557	0.872	4.805	Deep Video Super-Resolution using HR Optical Flow Estimation
SRMD	0.877	0.594	5.882	27.672	0	0.834	3.468	Learning a Single Convolutional Super-Resolution Network for Multiple Degradations
iSeeBetter	0.741	0.748	0.045	31.104	0.629	0.896	6.809	iSeeBetter: Spatio-Temporal Video Super Resolution using Recurrent-Generative Back-Projection Networks
RBPN	0.74	0.746	0.043	31.407	0.629	0.899	7.068	Recurrent Back-Projection Network for Video Super-Resolution
ESRGAN	0.948	0.735	1.004	27.33	0	0.808	5.353	ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks

0 of 32 row(s) selected.