Video Quality Assessment on CVD 2014 (test)
0.896SRCCFasterVQA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FasterVQAResolution Range=480P - 720P, Efficiency=4X than FAST-VQA2022.10 | 0.896 | 0.904 | |
| FAST-VQAGroup=Ours, Protocol=Finetune2022.07 | 0.891 | 0.903 | |
| FAST-VQAResolution Range=480P - 720P2022.10 | 0.891 | 0.903 | |
| FAST-VQA-MGroup=Ours, Protocol=Finetune2022.07 | 0.877 | 0.892 | |
| FAST-VQA-MResolution Range=480P - 720P2022.10 | 0.877 | 0.892 | |
| PVQTraining Setting=Mixed Databases2021.08 | 0.8737 | 0.8793 | |
| BVQA-TCSVT-2022Group=Existing Fixed Deep, Resolution Range=480P - 720P2022.10 | 0.872 | 0.869 | |
| FAST-VQA w/o VQ-representationsGroup=Ours, Protocol=Finetune2022.07 | 0.871 | 0.888 | |
| VSFAGroup=Existing Fixed Deep, Protocol=Finetune2022.07 | 0.87 | 0.868 | |
| VSFAGroup=Existing Fixed Deep, Resolution Range=480P - 720P2022.10 | 0.87 | 0.868 | |
| Full-res Swin-T featuresGroup=Baseline Features, Protocol=Finetune2022.07 | 0.868 | 0.87 | |
| Full-res Swin-TResolution Range=480P - 720P, Features=Full-resolution Swin-T2022.10 | 0.868 | 0.87 | |
| CNN+TLVQMGroup=Ensemble C+D, Protocol=Finetune2022.07 | 0.863 | 0.88 | |
| CNN+TLVQMGroup=Ensemble C+D, Resolution Range=480P - 720P2022.10 | 0.863 | 0.88 | |
| Li et al.Model Type=Supervised Pretraining2022.06 | 0.862 | 0.882 | |
| Li et al.Model Type=Supervised Pretraining2022.06 | 0.862 | 0.882 | |
| CONVIQTModel Type=Unsupervised Pretraining2022.06 | 0.858 | 0.837 | |
| CONVIQTModel Type=Unsupervised Pretraining2022.06 | 0.858 | 0.837 | |
| VSFAModel Type=Supervised Pretraining2022.06 | 0.85 | 0.869 | |
| VSFAModel Type=Supervised Pretraining2022.06 | 0.85 | 0.869 | |
| Proposed-LSTraining Setting=Mixed Databases, Scaling=Linear Re-scaling (LS)2021.08 | 0.8406 | 0.8519 | |
| Resnet-50Model Type=Supervised Pretraining, Pretrained=ImageNet2022.06 | 0.835 | 0.838 | |
| Resnet-50Model Type=Supervised Pretraining2022.06 | 0.835 | 0.838 | |
| MDTVSFATraining Setting=Mixed Databases2021.08 | 0.8326 | 0.8347 | |
| GSTVQAModel Type=Supervised Pretraining2022.06 | 0.831 | 0.844 | |
| GSTVQAModel Type=Supervised Pretraining2022.06 | 0.831 | 0.844 | |
| GST-VQAGroup=Existing Fixed Deep, Protocol=Finetune2022.07 | 0.831 | 0.844 | |
| GST-VQAGroup=Existing Fixed Deep, Resolution Range=480P - 720P2022.10 | 0.831 | 0.844 | |
| TLVQMGroup=Existing Classical, Protocol=Finetune2022.07 | 0.83 | 0.85 | |
| TLVQMGroup=Existing Classical, Resolution Range=480P - 720P2022.10 | 0.83 | 0.85 | |
| CONTRIQUEModel Type=Unsupervised Pretraining2022.06 | 0.826 | 0.836 | |
| CONTRIQUEModel Type=Unsupervised Pretraining2022.06 | 0.826 | 0.836 | |
| VIDEVALModel Type=Traditional/Handcrafted Features2022.06 | 0.814 | 0.832 | |
| VIDEVALModel Type=Traditional/Handcrafted Features2022.06 | 0.814 | 0.832 | |
| RAPIQUEModel Type=Traditional/Handcrafted Features2022.06 | 0.807 | 0.823 | |
| RAPIQUEModel Type=Traditional/Handcrafted Features2022.06 | 0.807 | 0.823 | |
| VBLIINDSModel Type=Traditional/Handcrafted Features2022.06 | 0.795 | 0.806 | |
| VBLIINDSModel Type=Traditional/Handcrafted Features2022.06 | 0.795 | 0.806 | |
| BRISQUEModel Type=Traditional/Handcrafted Features2022.06 | 0.79 | 0.804 | |
| BRISQUEModel Type=Traditional/Handcrafted Features2022.06 | 0.79 | 0.804 | |
| TLVQMModel Type=Traditional/Handcrafted Features2022.06 | 0.779 | 0.79 | |
| TLVQMModel Type=Traditional/Handcrafted Features2022.06 | 0.779 | 0.79 | |
| CSPTModel Type=Unsupervised Pretraining, Evaluation Protocol=linear evaluation2022.06 | 0.653 | 0.691 | |
| CSPTModel Type=Unsupervised Pretraining, Evaluation Protocol=Linear evaluation2022.06 | 0.653 | 0.691 |