Research map / Video / Video segmentation and tracking
Distilling Cross-Temporal Contexts for Continuous Sign Language Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use CTCA instead of RadialCTC | CSLR on RWTH-2014 · RWTH-2014 test | WER (%): 20.2 -> 20.1 | Helped |
| Use CTCA instead of C2SLR | CSLR on RWTH-2014T · RWTH-2014T test | WER (%): 20.4 -> 20.3 | Helped |
| Use CTCA instead of the strongest prior test result | CSLR on CSL-Daily · CSL-Daily test | WER (%): 32.0 -> 29.4 | Helped |
| Add the SPM-shared dual-path network (ablation) | Baseline CSLR · RWTH-2014 test | WER (%): 22.1 -> 21.8 | Helped |
| Add the SPM-shared dual-path network (ablation) | VAC(VE) · RWTH-2014 test | WER (%): 23.8 -> 22.6 | Helped |
| Use vanilla distillation (ablation) | Baseline CSLR · RWTH-2014 test | WER (%): 22.1 -> 21.9 | Helped |
| Add local-to-global Ll2g distillation (ablation) | SDPN · RWTH-2014 test | WER (%): 22.1 -> 21.1 | Helped |
| Remove intersection-information weighting from Ll2g (ablation) | SDPN A · RWTH-2014 test | WER (%): 21.1 -> 21.5 | Hurt |
| Add global-to-local Lg2l distillation (ablation) | SDPN · RWTH-2014 test | WER (%): 22.1 -> 20.7 | Helped |
| Add mutual Ll2g and Lg2l distillation (ablation) | SDPN · RWTH-2014 test | WER (%): 22.1 -> 20.6 | Helped |
| Add reconstruction loss Lrc (ablation) | SDPN with mutual temporal distillation · RWTH-2014 test | WER (%): 20.6 -> 20.4 | Helped |
| Remove entropy-based weighting from Lrc (ablation) | SDPN D · RWTH-2014 test | WER (%): 20.4 -> 20.6 | Hurt |
Closest papers by meaning
- Cross-Sentence Gloss Consistency for Continuous Sign Language Recognition (AAAI 2024 · similarity 0.89)
- 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 0.88)
- KD-MSLRT: Lightweight Sign Language Recognition Model Based on Mediapipe and 3D to 1D Knowledge Distillation (AAAI 2025 · similarity 0.87)
- SCOPE: Sign Language Contextual Processing with Embedding from LLMs (AAAI 2025 · similarity 0.87)
- Lost in Translation, Found in Context: Sign Language Translation with Contextual Cues (CVPR 2025 · similarity 0.86)
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.