In-Domain Isolated Sign Language Recognition Setting

Single-view RGB-based ISLR involves recognizing isolated sign language from video sequences captured from a single fixed camera.

Single-view RGB-D-based ISLR aims to enhance the recognition of isolated signs by incorporating depth information along with RGB data.

Multi-view RGB-based ISLR employs multiple cameras to capture the sign language videos.

Multi-view RGB-D-based ISLR incorporates depth data in a multi-view setup.

Baseline Models

We mention that all models used in this work are publicly available. We express profound gratitude to the aforementioned authors for their invaluable contributions. Each of the ISLR models we use is linked below:

• RGB-based & RGB-D-based model:
  • ResNet2+1D GitHub
  • TSN GitHub
  • I3D GitHub
  • S3D GitHub
  • SlowFast GitHub
  • Timesformer GitHub
  • UMDR GitHub
  • KVNet-V GitHub
• 2D pose-based & 3D pose-based model:
  • TGCN GitHub
  • SL-GCN GitHub
  • SPTOTER GitHub
  • KVNet-K GitHub
• Multi-modal-based model:
  • SAM-SLR GitHub
  • NLA-SLR GitHub