Isolated Sign Language Recognition on NMFs-CSL (Total)
83.6Top-1 AccStepNet (R + F)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| StepNet (R + F)Modality=Fusion-based, Input Details=RGB data (R) + Optical flow data (F)2022.12 | 83.6 | 92.7 | 97 | |
| BEST (H + R)Modality=Fusion-based, Input Details=H + RGB-based features (R)2022.12 | 79.2 | — | 97.1 | |
| SignBERT (H + R)Input Modality=RGB-based, Configuration=Hand + RGB (H+R)2021.10 | 78.4 | 92 | 97.3 | |
| Signbert (H + R)Modality=Fusion-based, Input Details=H + RGB-based features (R)2022.12 | 78.4 | 92 | 97.3 | |
| StepNetModality=RGB-based2022.12 | 77.2 | 86.2 | 92.5 | |
| SignBERT (H + P)Input Modality=Pose-based, Configuration=Hand + Pose (H+P)2021.10 | 74.9 | 93.2 | 98.2 | |
| Signbert (H + P)Modality=Fusion-based, Input Details=H + Skeleton-based features (P)2022.12 | 74.9 | 93.2 | 98.2 | |
| GLE-NetInput Modality=RGB-based2021.10 | 69 | 79.9 | 88.1 | |
| GLE-NetModality=RGB-based2022.12 | 69 | 79.9 | 88.1 | |
| BESTModality=Skeleton-based2022.12 | 68.5 | — | 94.4 | |
| SignBERT (H)Input Modality=Pose-based, Configuration=Hand pose (H)2021.10 | 67 | 86.8 | 95.3 | |
| Signbert (H)Modality=Skeleton-based, Input Details=Reconstructed hand representation (H) by MANO2022.12 | 67 | 86.8 | 95.3 | |
| SlowfastInput Modality=RGB-based2021.10 | 66.3 | 77.8 | 86.6 | |
| SlowfastModality=RGB-based2022.12 | 66.3 | 77.8 | 86.6 | |
| TSMInput Modality=RGB-based2021.10 | 64.5 | 79.5 | 88.7 | |
| TSMModality=RGB-based2022.12 | 64.5 | 79.5 | 88.7 | |
| I3DInput Modality=RGB-based2021.10 | 64.4 | 77.9 | 88 | |
| I3DModality=RGB-based2022.12 | 64.4 | 77.9 | 88 | |
| 3D-R50Input Modality=RGB-based2021.10 | 62.1 | 73.2 | 82.9 | |
| 3D-R50Modality=RGB-based2022.12 | 62.1 | 73.2 | 82.9 | |
| ST-GCNInput Modality=Pose-based2021.10 | 59.9 | 74.7 | 86.8 | |
| ST-GCNModality=Skeleton-based2022.12 | 59.9 | 74.7 | 86.8 | |
| DNFInput Modality=RGB-based2021.10 | 55.8 | 69.5 | 82.4 | |
| DNFModality=RGB-based2022.12 | 55.8 | 69.5 | 82.4 |