Video Quality Assessment on KoNViD 1k (test)
0.895SRCCFasterVQA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FasterVQAResolution Range=540P, Efficiency=4X than FAST-VQA2022.10 | 0.895 | — | 0.898 | |
| FAST-VQAGroup=Ours, Protocol=Finetune2022.07 | 0.891 | 0.892 | — | |
| FAST-VQAResolution Range=540P2022.10 | 0.891 | — | 0.892 | |
| DOVERTraining Set=LSVQ_train, GFLOPS=282.3, CPU Time=9.7s, GPU Time=47ms2022.11 | 0.884 | — | 0.883 | |
| FAST-VQA-MGroup=Ours, Protocol=Finetune2022.07 | 0.873 | 0.872 | — | |
| FAST-VQA-MResolution Range=540P2022.10 | 0.873 | — | 0.872 | |
| FAST-VQATraining Set=LSVQ_train, GFLOPS=279.1, CPU Time=8.8s, GPU Time=45ms2022.11 | 0.859 | — | 0.855 | |
| FAST-VQA w/o VQ-representationsGroup=Ours, Protocol=Finetune2022.07 | 0.842 | 0.844 | — | |
| Full-res Swin-T featuresGroup=Baseline Features, Protocol=Finetune2022.07 | 0.841 | 0.838 | — | |
| Full-res Swin-TResolution Range=540P, Features=Full-resolution Swin-T2022.10 | 0.841 | — | 0.838 | |
| ProposedTraining Dataset=LSVQ, Evaluation Protocol=Cross-database2021.08 | 0.839 | 0.83 | — | |
| BVQA-TCSVT-2022Group=Existing Fixed Deep, Resolution Range=540P2022.10 | 0.834 | — | 0.836 | |
| Li et al.Training Set=LSVQ_train, GFLOPS=112537, CPU Time=1567s, GPU Time=27.6s2022.11 | 0.834 | — | 0.837 | |
| PVQTraining Setting=Mixed Databases2021.08 | 0.8205 | 0.8142 | — | |
| CNN+TLVQMGroup=Ensemble C+D, Protocol=Finetune2022.07 | 0.816 | 0.818 | — | |
| CNN+TLVQMGroup=Ensemble C+D, Resolution Range=540P2022.10 | 0.816 | — | 0.818 | |
| CNN+VIDEVALGroup=Ensemble C+D, Protocol=Finetune2022.07 | 0.815 | 0.817 | — | |
| CNN+VIDEVALGroup=Ensemble C+D, Resolution Range=540P2022.10 | 0.815 | — | 0.817 | |
| GST-VQAGroup=Existing Fixed Deep, Protocol=Finetune2022.07 | 0.814 | 0.825 | — | |
| GST-VQAGroup=Existing Fixed Deep, Resolution Range=540P2022.10 | 0.814 | — | 0.825 | |
| RAPIQUEGroup=Existing Classical, Protocol=Finetune2022.07 | 0.803 | 0.817 | — | |
| RAPIQUEGroup=Existing Classical, Resolution Range=540P2022.10 | 0.803 | — | 0.817 | |
| Proposed-LSTraining Setting=Mixed Databases, Scaling=Linear Re-scaling (LS)2021.08 | 0.7976 | 0.7896 | — | |
| DisCoVQATrain on=LIVE-VQC2022.06 | 0.792 | 0.785 | — | |
| PVQtraining_dataset=LSVQ, fine-tuning=false, patch_mode=w/ v-patch2020.11 | 0.791 | 0.795 | — | |
| PVQ(w)Training Dataset=LSVQ, Evaluation Protocol=Cross-database2021.08 | 0.791 | 0.795 | — | |
| PVQGroup=Existing Fixed Deep, Protocol=Finetune2022.07 | 0.791 | 0.786 | — | |
| PVQGroup=Existing Fixed Deep, Resolution Range=540P2022.10 | 0.791 | — | 0.786 | |
| Patch-VQ (w/ patch)Training Set=LSVQ_train, GFLOPS=58501, CPU Time=539s, GPU Time=13.8s2022.11 | 0.791 | — | 0.795 | |
| VSFAtraining_dataset=LSVQ, fine-tuning=false2020.11 | 0.784 | 0.794 | — | |
| VSFATraining Dataset=LSVQ, Evaluation Protocol=Cross-database2021.08 | 0.784 | 0.794 | — | |
| VSFATraining Set=LSVQ_train, GFLOPS=40919, CPU Time=466s, GPU Time=11.1s2022.11 | 0.784 | — | 0.795 | |
| VIDEVALGroup=Existing Classical, Protocol=Finetune2022.07 | 0.783 | 0.78 | — | |
| VIDEVALGroup=Existing Classical, Resolution Range=540P2022.10 | 0.783 | — | 0.78 | |
| PVQtraining_dataset=LSVQ, fine-tuning=false, patch_mode=w/o v-patch2020.11 | 0.782 | 0.781 | — | |
| PVQ(w/o)Training Dataset=LSVQ, Evaluation Protocol=Cross-database2021.08 | 0.782 | 0.781 | — | |
| MDTVSFATraining Setting=Mixed Databases2021.08 | 0.7816 | 0.7786 | — | |
| Patch-VQ (w/o patch)Training Set=LSVQ_train, GFLOPS=58501, CPU Time=539s, GPU Time=13.8s2022.11 | 0.781 | — | 0.781 | |
| TLVQMGroup=Existing Classical, Protocol=Finetune2022.07 | 0.773 | 0.768 | — | |
| VSFAGroup=Existing Fixed Deep, Protocol=Finetune2022.07 | 0.773 | 0.775 | — | |
| TLVQMGroup=Existing Classical, Resolution Range=540P2022.10 | 0.773 | — | 0.768 | |
| VSFAGroup=Existing Fixed Deep, Resolution Range=540P2022.10 | 0.773 | — | 0.775 | |
| COINVQGroup=Existing Fixed Deep, Protocol=Finetune2022.07 | 0.767 | 0.764 | — | |
| CoINVQGroup=Existing Fixed Deep, Resolution Range=540P2022.10 | 0.767 | — | 0.764 | |
| VIDEVALtraining_dataset=LSVQ, fine-tuning=false2020.11 | 0.751 | 0.741 | — | |
| VIDEVALTraining Dataset=LSVQ, Evaluation Protocol=Cross-database2021.08 | 0.751 | 0.741 | — | |
| VIDEVALTraining Set=LSVQ_train, CPU Time=895s2022.11 | 0.751 | — | 0.741 | |
| TLVQMtraining_dataset=LSVQ, fine-tuning=false2020.11 | 0.732 | 0.724 | — | |
| TLVQMTraining Dataset=LSVQ, Evaluation Protocol=Cross-database2021.08 | 0.732 | 0.724 | — | |
| TLVQMTraining Set=LSVQ_train, CPU Time=248s2022.11 | 0.732 | — | 0.724 | |
| GST-VQATrain on=LIVE-VQC2022.06 | 0.709 | 0.707 | — | |
| MDTVSFATrain on=LIVE-VQC2022.06 | 0.706 | 0.711 | — | |
| DisCoVQATrain on=Youtube-UGC (236-fewer version)2022.06 | 0.686 | 0.697 | — | |
| KSVQETrain Dataset=KVQ2024.02 | 0.65 | — | 0.661 | |
| MDTVSFATrain on=Youtube-UGC2022.06 | 0.649 | 0.646 | — | |
| BRISQUEtraining_dataset=LSVQ, fine-tuning=false2020.11 | 0.646 | 0.647 | — | |
| BRISQUETraining Dataset=LSVQ, Evaluation Protocol=Cross-database2021.08 | 0.646 | 0.647 | — | |
| CNN-TLVQMTrain on=LIVE-VQC2022.06 | 0.642 | 0.631 | — | |
| FastVQATrain Dataset=KVQ2024.02 | 0.641 | — | 0.654 | |
| VIDEVALTrain on=LIVE-VQC2022.06 | 0.625 | 0.621 | — | |
| VIDEVALTrain on=Youtube-UGC2022.06 | 0.61 | 0.62 | — | |
| SimpleVQATrain Dataset=KVQ2024.02 | 0.475 | — | 0.481 |