Human Action Recognition on NTU RGB+D 120 Few-Shot
86.4Median Top-1 AccFukushi et al.
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Fukushi et al.Source=Prior Work2025.12 | 86.4 | — | |
| KineMICSource=Our Analysis2025.12 | 86.2 | — | |
| MDMSource=Our Analysis, Training strategy=pre-trained2025.12 | 83.9 | — | |
| Kinetic-GANSource=Prior Work2025.12 | 81.7 | — | |
| MDMSource=Our Analysis, Training strategy=LoRA fine-tune2025.12 | 75.5 | — | |
| MDMSource=Our Analysis, Training strategy=from scratch2025.12 | 73.9 | — | |
| ACTORSource=Prior Work2025.12 | 73.6 | — | |
| Real-data-onlySource=Our Analysis, Number of samples=302025.12 | 63.1 | — | |
| Real-data-onlySource=Prior Work, Number of samples=302025.12 | 58.4 | — | |
| PST-transformerShots=1-shot2026.06 | — | 54.55 | |
| PST-transformerShots=4-shot2026.06 | — | 70.58 | |
| PST-transformerShots=8-shot2026.06 | — | 76.68 | |
| SRENetShots=1-shot2026.06 | — | 68.53 | |
| SRENetShots=4-shot2026.06 | — | 82.76 | |
| SRENetShots=8-shot2026.06 | — | 87.15 | |
| SRENet (w/o re-entry)Shots=1-shot2026.06 | — | 66.62 | |
| SRENet (w/o re-entry)Shots=4-shot2026.06 | — | 80.8 | |
| SRENet (w/o re-entry)Shots=8-shot2026.06 | — | 85.06 |