List your company
ResearchFormerly WizWand

Video Quality Assessment on LIVE-Qualcomm (test)

0.8387SRCCPVQ
0.5405320.6179410.695350.772759Aug 19, 2021Oct 27, 2021Jan 5, 2022Mar 16, 2022May 24, 2022Aug 2, 2022Oct 11, 2022
Updated 8mo ago

Evaluation results

MethodLinks
PVQTraining Setting=Mixed Databases2021.080.83870.8617
Li et al.Model Type=Supervised Pretraining2022.060.8330.837
Li et al.Model Type=Supervised Pretraining2022.060.8330.837
FasterVQAResolution Range=1080P, Efficiency=4X than FAST-VQA2022.100.8260.844
FAST-VQAGroup=Ours, Protocol=Finetune2022.070.8190.851
FAST-VQAResolution Range=1080P2022.100.8190.851
BVQA-TCSVT-2022Group=Existing Fixed Deep, Resolution Range=1080P2022.100.8170.828
MDTVSFATraining Setting=Mixed Databases2021.080.81360.8291
CNN+TLVQMGroup=Ensemble C+D, Protocol=Finetune2022.070.810.833
CNN+TLVQMGroup=Ensemble C+D, Resolution Range=1080P2022.100.810.833
Proposed-LSTraining Setting=Mixed Databases, Scaling=Linear Re-scaling (LS)2021.080.80410.8253
FAST-VQA-MGroup=Ours, Protocol=Finetune2022.070.8040.838
FAST-VQA-MResolution Range=1080P2022.100.8040.838
GSTVQAModel Type=Supervised Pretraining2022.060.8010.825
GSTVQAModel Type=Supervised Pretraining2022.060.8010.825
GST-VQAGroup=Existing Fixed Deep, Protocol=Finetune2022.070.8010.825
GST-VQAGroup=Existing Fixed Deep, Resolution Range=1080P2022.100.8010.825
CONVIQTModel Type=Unsupervised Pretraining2022.060.7970.802
CONVIQTModel Type=Unsupervised Pretraining2022.060.7970.802
Full-res Swin-T featuresGroup=Baseline Features, Protocol=Finetune2022.070.7880.803
Full-res Swin-TResolution Range=1080P, Features=Full-resolution Swin-T2022.100.7880.803
TLVQMModel Type=Traditional/Handcrafted Features2022.060.7850.815
TLVQMModel Type=Traditional/Handcrafted Features2022.060.7850.815
TLVQMGroup=Existing Classical, Protocol=Finetune2022.070.770.81
TLVQMGroup=Existing Classical, Resolution Range=1080P2022.100.770.81
CONTRIQUEModel Type=Unsupervised Pretraining2022.060.7650.777
CONTRIQUEModel Type=Unsupervised Pretraining2022.060.7650.777
FAST-VQA w/o VQ-representationsGroup=Ours, Protocol=Finetune2022.070.7560.778
VSFAGroup=Existing Fixed Deep, Protocol=Finetune2022.070.7370.779
VSFAGroup=Existing Fixed Deep, Resolution Range=1080P2022.100.7370.732
VSFAModel Type=Supervised Pretraining2022.060.7080.774
VSFAModel Type=Supervised Pretraining2022.060.7080.774
VIDEVALModel Type=Traditional/Handcrafted Features2022.060.670.705
VIDEVALModel Type=Traditional/Handcrafted Features2022.060.670.705
RAPIQUEModel Type=Traditional/Handcrafted Features2022.060.6650.691
RAPIQUEModel Type=Traditional/Handcrafted Features2022.060.6650.691
Resnet-50Model Type=Supervised Pretraining, Pretrained=ImageNet2022.060.650.691
Resnet-50Model Type=Supervised Pretraining2022.060.650.691
VBLIINDSModel Type=Traditional/Handcrafted Features2022.060.570.626
VBLIINDSModel Type=Traditional/Handcrafted Features2022.060.570.626
BRISQUEModel Type=Traditional/Handcrafted Features2022.060.5520.598
BRISQUEModel Type=Traditional/Handcrafted Features2022.060.5520.598