List your company
ResearchFormerly WizWand

Video Quality Assessment on YouTube-UGC (test)

0.863SRCCFasterVQA
0.308680.452590.59650.74041May 29, 2020May 16, 2021May 3, 2022Apr 20, 2023Apr 6, 2024Mar 24, 2025Mar 12, 2026
Updated 6mo ago

Evaluation results

MethodLinks
FasterVQAResolution Range=360P - 2160P(4K), Efficiency=4X than FAST-VQA2022.100.8630.859
FAST-VQAGroup=Ours, Protocol=Finetune2022.070.8550.852
FAST-VQAResolution Range=360P - 2160P(4K)2022.100.8550.852
Q-AlignTraining Dataset=LSVQ, Fine-tuning=False2025.050.8330.846
BVQA-TCSVT-2022Group=Existing Fixed Deep, Resolution Range=360P - 2160P(4K)2022.100.8310.819
MDS-VQALabeling Budget=5%2026.030.8190.807
CoINVQGroup=Existing Fixed Deep, Resolution Range=360P - 2160P(4K)2022.100.8160.802
FreeSelLabeling Budget=5%2026.030.8140.798
CP-LLMTraining Dataset=LSVQ, Fine-tuning=False2025.050.810.813
CNN+VIDEVALGroup=Ensemble C+D, Protocol=Finetune2022.070.8080.803
CNN+VIDEVALGroup=Ensemble C+D, Resolution Range=360P - 2160P(4K)2022.100.8080.803
NoiseStabilityLabeling Budget=5%2026.030.8060.798
Full-res Swin-T featuresGroup=Baseline Features, Protocol=Finetune2022.070.7980.796
Full-res Swin-TResolution Range=360P - 2160P(4K), Features=Full-resolution Swin-T2022.100.7980.796
Sampling by RDLabeling Budget=5%2026.030.7950.784
FAST-VQA w/o VQ-representationsGroup=Ours, Protocol=Finetune2022.070.7940.784
ModularBVQATraining Dataset=LSVQ, Fine-tuning=False2025.050.7880.804
ALCSLabeling Budget=5%2026.030.7850.772
Greedy samplingLabeling Budget=5%2026.030.7810.773
VIDEVALGroup=Existing Classical, Resolution Range=360P - 2160P(4K)2022.100.7790.773
DOVERTraining Dataset=LSVQ, Fine-tuning=False2025.050.7770.792
Core-set selectionLabeling Budget=5%2026.030.7760.765
VIDEVALGroup=Existing Classical, Protocol=Finetune2022.070.7730.773
FAST-VQA-MGroup=Ours, Protocol=Finetune2022.070.7680.765
FAST-VQA-MResolution Range=360P - 2160P(4K)2022.100.7680.765
Random samplingLabeling Budget=5%2026.030.760.756
MC dropoutLabeling Budget=5%2026.030.760.753
RAPIQUEGroup=Existing Classical, Protocol=Finetune2022.070.7590.768
RAPIQUEGroup=Existing Classical, Resolution Range=360P - 2160P(4K)2022.100.7590.768
KSVQETrain Dataset=KVQ2024.020.7420.764
VSFAGroup=Existing Fixed Deep, Protocol=Finetune2022.070.7320.743
FastVQATraining Dataset=LSVQ, Fine-tuning=False2025.050.730.747
VSFAGroup=Existing Fixed Deep, Resolution Range=360P - 2160P(4K)2022.100.7240.743
Base modelLabeling Budget=5%2026.030.7080.709
SimpleVQATrain Dataset=KVQ2024.020.6750.674
TLVQMGroup=Existing Classical, Protocol=Finetune2022.070.6690.659
TLVQMGroup=Existing Classical, Resolution Range=360P - 2160P(4K)2022.100.6690.659
FastVQATrain Dataset=KVQ2024.020.6450.676
GST-VQAGroup=Existing Fixed Deep, Protocol=Finetune2022.070.610.592
VIDEVALTrain set=KoNViD-1k2020.050.37—
ResNet-50Train set=LIVE-VQC2020.050.33—