Multi-View Detection on WildTrack
93.6MODAPOV-MVD
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| POV-MVDTraining Mode=Single scene (Direct training)2026.03 | 93.6 | 82.4 | 96.6 | 97 | 96.8 | |
| 3DROMdistance threshold=0.5m2024.09 | 93.5 | 75.9 | 97.2 | 96.2 | 96.7 | |
| 3DROMTraining Mode=Single scene (Direct training)2026.03 | 93.5 | 75.9 | 97.2 | 96.2 | 96.7 | |
| M-MVOTdistance metric=Mahalanobis, distance threshold=0.5m2024.09 | 92.1 | 81.3 | 94.5 | 97.8 | 96.1 | |
| MVOTTraining Mode=Single scene (Direct training)2026.03 | 92.1 | 81.3 | 94.5 | 97.8 | 96.1 | |
| TrackTacularTraining Mode=Single scene (Direct training)2026.03 | 92.1 | 76.2 | 97 | 95.1 | 96 | |
| E-MVOTdistance metric=Euclidean, distance threshold=0.5m2024.09 | 91.9 | 80.9 | 94.1 | 98 | 96 | |
| MVDeTrdistance threshold=0.5m2024.09 | 91.5 | 82.1 | 97.4 | 94 | 95.7 | |
| MVDeTrTraining Mode=Single scene (Direct training)2026.03 | 91.5 | 82.1 | 97.4 | 94 | 95.7 | |
| EarlyBirdTraining Mode=Single scene (Direct training)2026.03 | 91.2 | 81.8 | 94.9 | 96.3 | 95.6 | |
| SHOTdistance threshold=0.5m2024.09 | 90.2 | 76.5 | 96.1 | 94 | 95 | |
| SHOTTraining Mode=Single scene (Direct training)2026.03 | 90.2 | 76.5 | 96.1 | 94 | 95 | |
| Baseline (100%+DA)Training data percentage=100%, Domain adaptation (DA)=true, Pre-training=SynMVCrowd2026.03 | 88.7 | 81.3 | 95.3 | 93.9 | 94.6 | |
| Volumetricdistance threshold=0.5m2024.09 | 88.6 | 73.8 | 95.3 | 93.2 | 94.2 | |
| VolumetricTraining Mode=Single scene (Direct training)2026.03 | 88.6 | 73.8 | 95.3 | 93.2 | 94.2 | |
| MVDetdistance threshold=0.5m2024.09 | 88.2 | 75.7 | 94.7 | 93.6 | 94.1 | |
| MVDetTraining Mode=Single scene (Direct training)2026.03 | 88.2 | 75.7 | 94.7 | 93.6 | 94.1 | |
| Baseline (100%)Training data percentage=100%, Domain adaptation (DA)=false, Pre-training=SynMVCrowd2026.03 | 85.9 | 76.5 | 94.1 | 91.7 | 92.9 | |
| Ours (DropView)Trained on=GMVD training set2021.09 | 80.1 | 75.6 | 90.9 | 89.1 | — | |
| Baseline (30%+DA)Training data percentage=30%, Domain adaptation (DA)=true, Pre-training=SynMVCrowd2026.03 | 75.3 | 78.4 | 89.7 | 89.2 | 89.4 | |
| DeepOcc.distance threshold=0.5m2024.09 | 74.1 | 53.8 | 95 | 80 | 86.9 | |
| DeepOcc.Training Mode=Single scene (Direct training)2026.03 | 74.1 | 53.8 | 95 | 80 | 86.9 | |
| DeepMCDdistance threshold=0.5m2024.09 | 67.8 | 64.2 | 85 | 82 | 83.5 | |
| DeepMCDTraining Mode=Single scene (Direct training)2026.03 | 67.8 | 64.2 | 85 | 82 | 83.5 | |
| Baseline (30%)Training data percentage=30%, Domain adaptation (DA)=false, Pre-training=SynMVCrowd2026.03 | 67.8 | 71.1 | 83.2 | 84.9 | 84 | |
| Baseline (10%+DA)Training data percentage=10%, Domain adaptation (DA)=true, Pre-training=SynMVCrowd2026.03 | 63.2 | 73.6 | 82.4 | 83.9 | 83.2 | |
| Baseline (10%)Training data percentage=10%, Domain adaptation (DA)=false, Pre-training=SynMVCrowd2026.03 | 55 | 68.3 | 77.6 | 77.3 | 77.5 | |
| Baseline (test)Training data percentage=0% (Direct Transfer), Domain adaptation (DA)=false, Pre-training=SynMVCrowd2026.03 | 31.9 | 48.5 | 78.4 | 44.1 | 56.5 | |
| POM-CNNdistance threshold=0.5m2024.09 | 23.2 | 30.5 | 75 | 55 | 63.5 | |
| POM-CNNTraining Mode=Single scene (Direct training)2026.03 | 23.2 | 30.5 | 75 | 55 | 63.5 | |
| RCNNdistance threshold=0.5m2024.09 | 11.3 | 18.4 | 68 | 43 | 52.7 | |
| RCNNTraining Mode=Single scene (Direct training)2026.03 | 11.3 | 18.4 | 68 | 43 | 52.7 |