Action Spotting on SoccerNet v2 (test)
73.74Average-mAP (Tight 1-5 s)Proposed (ASTRM/SoftIC)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Proposed (ASTRM/SoftIC)Params (M)=6.46, GFLOPS=60.252025.02 | 73.74 | 79.11 | — | |
| COMEDIANBackbone=ViSwin-T, Params (M)=70.12, GFLOPS=222.762025.02 | 71.33 | 77.07 | — | |
| ASTRAParams (M)=44.33, GFLOPS=8.832025.02 | 66.63 | 78.14 | — | |
| DU+SAM+mixup +Soft-NMSFeatures=Combination×2 + ResNet2022.06 | 65.1 | — | 78.5 | |
| SpivakParams (M)=17.46, GFLOPS=461.892025.02 | 65.1 | 78.5 | — | |
| UGLFParams (M)=4.46, GFLOPS=39.612025.02 | 63.51 | 73.98 | — | |
| E2E-SpotBackbone=800MF2022.07 | 61.82 | 74.05 | — | |
| E2E-SpotBackbone=RegNet-Y 800MF, Params (M)=12.64, GFLOPS=151.42025.02 | 61.82 | 74.05 | — | |
| E2E-SpotFeatures=RegNetY tuned2022.06 | 61.8 | — | 74.1 | |
| E2E-SpotBackbone=200MF2022.07 | 61.19 | 73.25 | — | |
| E2E-SpotBackbone=RegNet-Y 200MF, Params (M)=4.46, GFLOPS=39.612025.02 | 61.19 | 73.25 | — | |
| SpotFormer2025.02 | 60.9 | 76.1 | — | |
| DU+SAM+mixupFeatures=Combination2022.06 | 60.7 | — | 77.3 | |
| Faster-TADFeatures=Swin tuned2022.06 | 54.1 | — | — | |
| Zhou et al.Features=Combination2022.06 | 47.1 | — | 73.8 | |
| Zhou et al.Competition Context=2021 challenge; 1st2022.07 | 47.05 | 73.77 | — | |
| Baidu (TF)Backbone=TF2025.02 | 47.05 | 73.77 | — | |
| T-DEEDBackbone=RegNet-Y 200MF, Params (M)=16.36, GFLOPS=21.962025.02 | 39.43 | 52.93 | — | |
| RMS-Net2022.07 | 28.83 | 63.49 | — | |
| AImageLab RMSNetFeatures=ResNet tuned2022.06 | 28.8 | — | 63.5 | |
| T-DEEDBackbone=RegNet-Y 800MF, Params (M)=46.22, GFLOPS=60.252025.02 | 21.57 | 30.49 | — | |
| CALFFeatures=ResNet+PCA2022.06 | 12.2 | — | 40.7 | |
| NetVLAD++Features=ResNet2022.06 | 11.5 | — | 53.4 | |
| AlmageLab RMSNetInput type=Feature extractor2023.09 | — | — | 28.8 | |
| BaiduInput type=Feature extractor2023.09 | — | — | 47.1 | |
| CALF-calibrationFeatures=ResNet+PCA + top view2022.06 | — | — | 46.8 | |
| COMEDIANInput type=End-to-end (Image), Backbone=ViViT-T2023.09 | — | — | 70.7 | |
| COMEDIANInput type=End-to-end (Image), Backbone=ViSwin-T2023.09 | — | — | 71.6 | |
| COMEDIANInput type=End-to-end (Image), Backbone=ViViT-T, Ensemble=true2023.09 | — | — | 72 | |
| COMEDIANInput type=End-to-end (Image), Backbone=ViSwin-T, Ensemble=true2023.09 | — | — | 73.1 | |
| E2E-SpotInput type=End-to-end (Image)2023.09 | — | — | 61.8 | |
| Faster-TADInput type=Feature extractor2023.09 | — | — | 54.1 | |
| NetVLAD++Input type=Feature extractor2023.09 | — | — | 11.5 | |
| SpivakInput type=Feature extractor2023.09 | — | — | 65.1 | |
| SpotFormerInput type=Feature extractor2023.09 | — | — | 60.9 |