%0 Computer Program %A Butler, Rick %D 2025 %T Python package to read 2D human poses from json files %U %R 10.4121/97d8618d-d456-4c13-9a6c-637217643fdb.v1 %K 2D pose detection %K Data loading %K Multiprocessing %K Pose detection/tracklet conversion %X
About 200 coronary angiogram procedures were recorded from a distance in the Reinier de Graaf hospital, Delft, NL.
Human pose data were extracted from these videos and saved.
As the dataset is large, a tool was needed to read the pose data for usage.
This Python package reads poses from .det2d.json files, and converts between detection and tracklet formats.
For large datasets, it offers functions to read pose data one file at a time such that the entire dataset need not be loaded into memory.
It uses multiprocessing to buffer poses in memory for faster access.
%I 4TU.ResearchData