The project
Research and showcase by Preet Prasad, University of Western Australia Honours, supervised by Dr Atif Mansoor. The site presents ongoing research; no endorsement by dataset authors, programme creators or performers is implied.
Project code and its licence are available in the research repository. The code licence does not replace separate licences for datasets, model weights or third-party dependencies.
Recorded signing examples
The four example clips are from the official validation split of Auslan-Daily: Australian Sign Language Translation for Daily Communication and News, by Xin Shen, Shaozu Yuan, Hongwei Sheng, Heming Du and Xin Yu (NeurIPS 2023). The conversational material includes the programme Sally & Possum.
The current dataset download page states Creative Commons Attribution 4.0 International, checked on 25 September 2026. The showcase relies on that statement. The 2023 paper’s footnote 8 instead names CC BY-NC-ND 4.0; these sources are not consistent. Consult the dataset maintainers before relying on this page for a different reuse.
This is a noncommercial academic research showcase. The linked licence gives no warranties, and other rights may apply.
What was changed
The display videos are silent H.264 copies for browser compatibility, with no additional crop. Source frame counts and the 25-frame-per-second timing are preserved. Each poster is the first source frame.
DWPose tracking, normalised coordinates, encoder features, self-attention and model predictions were produced by this project. Overlays are drawn separately in the browser. These added outputs are not annotations endorsed by the dataset authors.
Each manifest records source and display hashes, processing and model provenance. The downloadable attribution record provides further details.
Research components
The research uses or evaluates DWPose movement tracking, Transformer translation models and, in image-based systems, DINOv2 image features. Isolated-sign material used in training experiments comes from MM-WLAuslan. See the repository for versioned dependencies, component licences and experiment configurations.
Browser speech uses an available local English voice supplied by the visitor’s device. Voice availability varies; research timing measurements used macOS Karen.
Credit or rights concerns
If an attribution is incomplete or you have a concern about material shown here, contact 23857493@student.uwa.edu.au with the page URL and clip identifier. Please describe the correction or action you are requesting.