The CogVis Lab develops vision and language models that understand, translate, and produce sign languages — bridging communication between Deaf and hearing communities through gloss-free translation, avatar-driven production, and LLM-powered annotation.
We work at the intersection of computer vision, natural language processing, and accessibility to build systems that understand and generate sign languages.
Gloss-free, end-to-end translation of continuous sign language video into spoken language text using visual transformers and large language models.
Generating natural, diverse sign language video from spoken text via motion tokenisation, sparse keyframe learning, and photorealistic avatar synthesis.
Detecting and disambiguating individual signs in unconstrained continuous video, including fingerspelling and rare or out-of-vocabulary signs.
Leveraging agentic large language models for linguistically grounded sign annotation, dataset curation, and multilingual sign language understanding.
We publish at top venues in computer vision and machine learning. See all on Google Scholar →
Harry Walsh and Ryan Wong has graduated with PhDs! Congratulations to Harry and Ryan on successfully defending their theses and earning their doctorates.
SignGPT project launched! We are excited to announce the launch of SignGPT, our new initiative focused on leveraging large language models for sign language understanding and generation. Stay tuned for updates on our research and developments in this area.