Related Experiment Video
Updated: Jan 17, 2026
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
1.0K
SignVLM: a pre-trained large video model for sign language recognition
1Information and Computer Science Department, King Fahd University of Petroleum and Minerals, SDAIA-KFUPM Joint Research Center for Artificial Intelligence, Saudi Arabia.
Peerj. Computer Science
|September 24, 2025
Summary
Sign language recognition (SLR) is improved by SignVLM, a pretrained vision model. SignVLM effectively handles low-resource sign languages, demonstrating strong performance across multiple datasets.
Area of Science:
- Computer Science
- Artificial Intelligence
- Human-Computer Interaction
Background:
- Sign language recognition (SLR) is crucial for community inclusion of individuals with hearing impairments.
- Developing effective SLR systems is hindered by a scarcity of annotated datasets, particularly for low-resourced sign languages.
Purpose of the Study:
- To introduce SignVLM, a novel pretrained large vision model designed to enhance sign language recognition.
- To investigate the efficacy of the contrastive language-image pre-training (CLIP) model for SLR tasks.
Main Methods:
- Utilized a pretrained CLIP model to extract spatial features from sign video frames.
- Employed a Transformer decoder for temporal learning within the SLR system.
- Evaluated the SignVLM model on four diverse sign language datasets: KArSL, WLASL, LSA64, and AUTSL.
Main Results:
- The proposed SignVLM model surpassed existing methods on the KArSL, WLASL, and LSA64 datasets.
- Achieved competitive performance on the AUTSL dataset, indicating robustness.
- Demonstrated effective generalization capabilities to new datasets with limited samples.
Conclusions:
- SignVLM shows significant promise for advancing SLR, especially in low-resource scenarios.
- The model's performance highlights the potential of large vision models in sign language processing.
- The findings support the adaptability and effectiveness of SignVLM for diverse sign language recognition applications.
