Computer vision to automatically assess infant neuromotor risk

Computer vision to automatically assess infant neuromotor risk

Claire Chambers, Nidhi Seethapathi, Rachit Saluja, Helen Loeb, Samuel Pierce, Daniel Bogen, Laura Prosser, Michelle J. Johnson, and Konrad Paul Kording

Code and instructions: https://github.com/cchamber/Infant_movement_assessment

This data set contains:

- System to extract pose estimates from video. (pose_extraction.zip)

- Zip file with 'data' and 'models' folders to add to code repository (infant_movement_assessment_repo_files.zip)

- Meta-data of YouTube videos in normative database of healthy infant movement (meta_data_youtube_dataset.csv)

- Meta-data of at-risk infant videos recorded in laboratory (meta_data_clinical_dataset.csv)

- Raw pose estimates from YouTube videos included in normative database of healthy infant movement (pose_estimates_youtube_dataset.csv)

- Raw pose estimates from videos recorded in laboratory (pose_estimates_clinical_dataset.csv)

- URLs of YouTube videos used to adapt pose-estimation algorithm (URL_labelled_dataset.csv)

- URLs of YouTube videos used for normative database of healthy infant movement (URL_pose_dataset.csv)

- Search terms used to find infant videos on YouTube for adapting OpenPose and to include in the normative database of healthy infant movement (search_terms.txt)

- Visualization of kinematic features

- Supplemental table describing infants tested in the laboratory