Multi-label object-presence prediction on CARRADA (test)
66.7Macro APMetricLearning-scene
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| MetricLearning-sceneEvaluation protocol=Frozen encoder with class-balanced logistic probe2026.06 | 66.7 | 0.044 | 90.5 | 18 | 84.4 | 0.18 | 0.076 | |
| FlatMLPEvaluation protocol=Frozen encoder with class-balanced logistic probe2026.06 | 66.1 | 0.011 | 94.6 | 12 | 95.3 | 0.004 | 0.004 | |
| OQ-TSAErawEvaluation protocol=Frozen encoder with class-balanced logistic probe, Variant=Reconstruction-only2026.06 | 65.7 | 0.01 | 91.1 | 11.6 | 96.9 | 0.032 | 0.018 | |
| TuckerAEEvaluation protocol=Frozen encoder with class-balanced logistic probe2026.06 | 65.6 | 0.012 | 89.1 | 9.5 | 94.6 | 0.03 | 0.014 | |
| ContrastiveSceneEvaluation protocol=Frozen encoder with class-balanced logistic probe2026.06 | 62.2 | 0.005 | 89.2 | 14 | 85.1 | 0.232 | 0.186 |