High-level Activity Labeling on CAD-120
84.7Micro P/Rfull model
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| full modelTemporal segmentation=Ground-truth2012.10 | 84.7 | 85.3 | 84.2 | |
| no temporal interactionsTemporal segmentation=Ground-truth2012.10 | 81.4 | 83.2 | 80.8 | |
| no object interactionsTemporal segmentation=Ground-truth2012.10 | 80.6 | 81.9 | 80 | |
| full, multi-seg learningTemporal segmentation=None (Estimated)2012.10 | 80.6 | 81.1 | 80 | |
| full model with trackingTemporal segmentation=Ground-truth2012.10 | 79 | 78.6 | 78.3 | |
| full, 1 segment. (averaged)Temporal segmentation=None (Estimated)2012.10 | 79 | 81.8 | 78.3 | |
| full, 1 segment. (best)Temporal segmentation=None (Estimated)2012.10 | 77.5 | 80.1 | 76.7 | |
| full, multi-seg learning + trackingTemporal segmentation=None (Estimated)2012.10 | 75 | 75.8 | 74.2 | |
| object onlyTemporal segmentation=Ground-truth2012.10 | 59.7 | 56.3 | 58.3 | |
| image onlyTemporal segmentation=Ground-truth2012.10 | 34.7 | 24.2 | 35.8 | |
| SVM multiclassTemporal segmentation=Ground-truth2012.10 | 30.6 | 27.4 | 31.2 | |
| sub-activity onlyTemporal segmentation=Ground-truth2012.10 | 27.4 | 31.8 | 27.7 | |
| MEMMTemporal segmentation=Ground-truth2012.10 | 26.4 | 23.7 | 23.7 | |
| max classTemporal segmentation=Ground-truth2012.10 | 10 | 10 | 10 |