Research map / Video / Video segmentation and tracking
Fully Convolutional Networks for Continuous Sign Language Recognition
Ka Leong Cheng, Zhaoyang Yang, Qifeng Chen, Yu-Wing Tai
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add GFE module with balance ratio (ablation) | FCN 1D-CNN/1D-CNN · RWTH test | WER (%): 26.0 -> 23.9 | Improved |
| add GFE module with balance ratio (ablation) | FCN 1D-CNN/1D-CNN · CSL test | WER (%): 8.2 -> 3.0 | Improved |
| add GFE module without balance ratio (ablation) | FCN 1D-CNN/1D-CNN · RWTH test | WER (%): 26.0 -> 25.4 | Improved |
| add GFE module without balance ratio (ablation) | FCN 1D-CNN/1D-CNN · CSL test | WER (%): 8.2 -> 4.5 | Improved |
| add balance ratio to GFE loss (ablation) | GFE module · RWTH test | WER (%): 25.4 -> 23.9 | Improved |
| add balance ratio to GFE loss (ablation) | GFE module · CSL test | WER (%): 4.5 -> 3.0 | Improved |
| add 1st-level 1D-CNN gloss encoder (ablation) | 2nd level 1D-CNN only · RWTH test | WER (%): 60.5 -> 26.0 | Improved |
| add 2nd-level 1D-CNN gloss encoder (ablation) | 1st level 1D-CNN only · RWTH test | WER (%): 42.1 -> 26.0 | Improved |
| replace LSTM 1st level with 1D-CNN (ablation) | LSTM/1D-CNN · RWTH test | WER (%): 32.1 -> 26.0 | Improved |
| replace 1D-CNN with BiLSTM at 2nd level (ablation) | 1D-CNN/1D-CNN · CSL test | WER (%): 8.2 -> 3.4 | Improved |
| replace 1D-CNN with BiLSTM at 2nd level (ablation) | 1D-CNN/1D-CNN · RWTH test | WER (%): 26.0 -> 26.5 | Hurt |
| use FCN instead of LSTM on shuffled glosses | LSTM-based network · RWTH test (Shuffle) | WER (%): 55.2 -> 28.1 | Improved |
Closest papers by meaning
- Recurrent Convolutional Neural Networks for Continuous Sign Language Recognition by Staged Optimization (CVPR 2017 · similarity 0.94)
- Video-Based Sign Language Recognition Without Temporal Segmentation (AAAI 2018 · similarity 0.91)
- Iterative Alignment Network for Continuous Sign Language Recognition (CVPR 2019 · similarity 0.91)
- Deep Sign: Enabling Robust Statistical Continuous Sign Language Recognition via Hybrid CNN-HMMs (IJCV 2018 · similarity 0.90)
- TCNet: Continuous Sign Language Recognition from Trajectories and Correlated Regions (AAAI 2024 · similarity 0.89)
- SignGraph: A Sign Sequence is Worth Graphs of Nodes (CVPR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
