MEDIAPI-SKEL
MEDIAPI-SKEL is a 2D-skeleton video corpus of LSFFrench Sign Language with French subtitles. The corpus consists of 368 subtitled videos produced by Média'Pi 
, a media company producing bilingual content with LSFFrench Sign Language and written French.The corpus was produced at the Laboratoire d'informatique pour la mécanique et les sciences de l'ingénieur (LIMSI).
From the original 368 videos 135 body keypoints were extracted from every signer in every frame using OpenPose. Version 2 of the dataset also adds a second keypoint model based on MediaPipe Holistic The subtitles are time aligned to the 2D-skeleton video content. Skeleton videos in JSON format and subtitles in WebVTT format are openly available, for the original videos an agreement with Média'Pi is needed.
| Language | French Sign Language |
|---|---|
| Size | 27 hours recorded, 17000 tokens from subtitles |
| Participants | more than 100 signers |
| Metadata Format | information not available |
| Translation | Fully translated (subtitles) (100%) |
| Annotation | Time aligned subtitles |
| Data Format | Open Pose |
| Licence |
Skeleton videos and subtitles: CC BY-NC 4.0 (an additonal note limits the use to non-commercial scientific research and teaching and redistribution to the the Ortolang platform).
Original videos: Upon written agreement with Média'Pi |
| Access |
Restricted access to video recordings for consultation only via Ortolang requires registration
Further access requires individual license agreement |
| Webpage | Dataset: https://hdl.handle.net/11403/mediapi-skel/v2 |
| Institution | Laboratoire d'informatique pour la mécanique et les sciences de l'ingénieur (LIMSI) |
Cite as
Bull, H., Braffort, A. and Gouiffès, M. (2020). MEDIAPI-SKEL - A 2D-Skeleton Video Database of French Sign Language With Aligned French Subtitles. In Proceedings of the Twelfth International Conference on Language Resources and Evaluation (LREC'2020), Marseille, France, May.
Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI) (2022). MEDIAPI-SKEL [Corpus]. ORTOLANG (Open Resources and TOols for LANGuage) - www.ortolang.fr, v1, https://hdl.handle.net/11403/mediapi-skel/v1.
This entry was last modified on 27 March 2025.