Behavior Classification on FMR1
76.3AccuracyGEESE
Evaluation results
| Method | Links | |
|---|---|---|
| GEESERepresentation=Fine-tuned embedding, Evaluation Protocol=End-to-End Fine-tuned2026.05 | 76.3 | |
| s-DANNCERepresentation=Wavelet + PCA (421-dim), Evaluation Protocol=Hand-crafted Features + Linear Probe2026.05 | 74.38 | |
| CEBRARepresentation=Contrastive embedding, Evaluation Protocol=Learned Representations + Linear Probe2026.05 | 59.84 | |
| PCA + SVMRepresentation=PCA components, Evaluation Protocol=Hand-crafted Features + Linear Probe2026.05 | 52.92 | |
| Raw pose + SVMRepresentation=Joint coordinates, Evaluation Protocol=Hand-crafted Features + Linear Probe2026.05 | 52.7 | |
| MOMENT (frozen)Representation=Pretrained embedding, Evaluation Protocol=Learned Representations + Linear Probe2026.05 | 49.75 | |
| Random GuessEvaluation Protocol=Reference2026.05 | 11.11 |