Word-level sign language recognition on WLASL 2000
81Top-1 AccSelf-supervised pre-trainable framework
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Self-supervised pre-trainable frameworkComputing Platform=Self-supervised pre-trainable framework with model aware hand prior incorporated.2026.01 | 81 | — | — | |
| Light-weight transformer-based modelComputing Platform=Light-weight transformer-based model, trained on 2D/3D keypoints.2026.01 | 69 | — | — | |
| Pose-based temporal graph convolution networksComputing Platform=Pose-based temporal graph convolution networks and comparison with an inflated 3D convolutional network.2026.01 | 64 | — | — | |
| Two-stream multistage graph convolutional modelComputing Platform=A two-stream multistage graph convolutional model with attention and residual connections designed to extract spatiotemporal contextual information.2026.01 | 63 | — | — | |
| I3D2019.10 | 32.48 | 57.31 | 66.31 | |
| Pose-TGCN2019.10 | 23.65 | 51.75 | 62.24 | |
| Pose-GRU2019.10 | 22.54 | 49.81 | 61.38 | |
| VGG-GRU2019.10 | 8.44 | 23.58 | 32.58 |