accessible-technology-in-sign / copycat-data-chi Goto Github PK

Dataset used in the CHI 2021 Student Research Competition Submission "CopyCat: Using Sign Language Recognition to Help Deaf Children Acquire Language Skills". This dataset contains AlphaPose, Kinect, and Mediapipe features extracted from 3914 4K RGB+D videos of ASL phrases and is intended for use by the interactive game CopyCat. In total, 8 users (novice adult ASL signers) are included and can be identified as p1 ... p8. In addition, 58 phrases ranging from 3 to 5 words in length were recorded at least 6 times for each user. All data is of the form MM-DD-YY_p(user id)_4KDepth/(phrase)/(instance of phrase)/MM-DD-YY_p(user_id)_4K.(phrase).(instance of phrase).(file extension) For further details, please refer to the paper/poster, the ASL recognition toolkit system ASLRT (ASL Recognition Toolkit), or the ASL recognition data generation system.