AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

BEST: BERT Pre-training for Sign Language Recognition with Coupling Tokenization

Weichao Zhao, Hezhen Hu, Wengang Zhou, Jiaxin Shi, Houqiang Li

AAAI 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Multilingual language representation pretraining · 55 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use coupled hand-and-body tokenization rather than separate vector quantizers (ablation)BEST · MSASL100Top-1 per-instance accuracy: 77.54 -> 80.98Improved
Use coupled hand-and-body tokenization rather than K-Means (ablation)BEST · MSASL100Top-1 per-instance accuracy: 74.63 -> 80.98Improved
Add hand-and-body masking instead of no pre-training (ablation)BEST · MSASL100Top-1 per-instance accuracy: 73.44 -> 80.98Improved
Mask hand features only rather than body features only (ablation)BEST · MSASL100Top-1 per-instance accuracy: 68.82 -> 76.75Improved
Add simultaneous hand-and-body masking rather than hand-only masking (ablation)BEST · MSASL100Top-1 per-instance accuracy: 76.75 -> 80.98Improved
Increase the fraction of pre-training data (ablation)BEST · MSASL100Top-1 per-instance accuracy: 74.50 -> 77.54 -> 78.47 -> 80.98Improved
Use MUM rather than the SignBERT random masking strategy (ablation)BEST pre-training · MSASL1000Top-1 per-instance accuracy: 53.95 -> 58.82Improved
Use token classification rather than regression as the pre-training objective (ablation)BEST pre-training · MSASL1000Top-1 per-instance accuracy: 56.90 -> 58.82Improved
Add BEST pre-training rather than train without pre-training (ablation)BEST · WLASL2000Top-1 per-instance accuracy: 27.13 -> 46.25Improved
Fuse BEST pose predictions with RGB predictionsBEST · MSASL100Top-1 per-instance accuracy: 80.98 -> 89.56Improved
Use BEST instead of SignBERTPose-based sign recognition · MSASL1000Top-1 per-instance accuracy: 49.54 -> 58.82Improved
Use BEST instead of SignBERTPose-based sign recognition · WLASL2000Top-1 per-instance accuracy: 39.40 -> 46.25Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy