BEAT
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Co-speech gesture generation | BEAT All Speakers 2 | BC8.103 | 31 | |
| Gesture Generation | BEAT2 | FGD28.15 | 17 | |
| Co-speech gesture generation | BEAT-2 (1 Speaker) | BC7.971 | 17 | |
| 3D co-speech gesture generation | BEAT-ETrans (test) | FGD (h+t)40.95 | 14 | |
| Co-speech gesture generation | BEAT | FGD4.247 | 13 | |
| Co-speech gesture generation | BEAT One Speaker v2 (Speaker 2) | FGD12.11 | 12 | |
| Gesture Generation | BEAT (test) | BC78.1 | 12 | |
| Holistic Motion Generation | BEAT2 | FGD4.623 | 12 | |
| Gesture Synthesis | BEAT2 Single-speaker (Scott) | BeatAlign0.795 | 9 | |
| Speech-driven Holistic Expression and Gesture Generation | BEAT 2022 (test) | FMD324.67 | 9 | |
| Co-speech Gesture Generation | BEAT 2 | Naturalness91.67 | 8 | |
| 3D Gesture Generation | BEAT (test) | BA0.83 | 8 | |
| Audio-driven Motion Generation | BEAT (test) | FID0 | 7 | |
| Speech-driven gesture generation | BEAT (test) | Global CCA100 | 7 | |
| Gesture Generation | BEAT official recomputed (test) | Hellinger Distance (Avg)0 | 7 | |
| Gesture Generation | BEAT (val) | FGD (Raw)0 | 6 | |
| Co-speech gesture generation | BEAT 2 (Novel 5 speakers) | BC68.6 | 6 | |
| Non-facial Gesture Generation | BEAT2 | FGD4.623 | 6 | |
| Text-to-Gesture Retrieval | BEAT2 N=856 (test) | Recall@142.3 | 5 | |
| Gesture-to-Text Retrieval | BEAT2 N=856 (test) | R@141.8 | 5 | |
| Co-speech gesture generation | BEAT2 All 25 Speakers | FGD6.145 | 5 | |
| Co-speech Gesture Synthesis | BEAT | Human-likeness4.61 | 5 | |
| Quantifier Prediction | BEAT2 (test) | Accuracy (%)74.8 | 5 | |
| Lip-synchronization | BEAT | MSE0.1185 | 5 | |
| Talking Head Generation | BEAT v2 (test) | MSE7.68 | 5 |
Showing 25 of 35 rows