Systems and methods for operator motion management

A computer-implemented method performed by an operator management module, the method comprising:

• obtaining video capture data captured by one or more capture devices, the video capture data comprising a plurality of images individually depicting a two-dimensional representation of a subject during performance of at least one activity of a set of activities to be performed by the subject within a workspace;
• identifying a set of points of the subject during performance of the at least one activity based at least in part on providing, to a machine-learning model as input, a subset of images from the video capture data, the machine-learning model being previously trained to identify points on a plurality of subjects in two-dimensional space from at least one image provided as input, the machine-learning model being trained utilizing a supervised learning algorithm and one or more historical images for which a corresponding set of points of a respective subject depicted in the historical image is known;
• updating a three-dimensional model depicting the subject during performance of the at least one activity, the three-dimensional model being updated based at least in part on the set of points identified by the machine-learning model;
• calculating, over a period of time, instances of kinematic data related to the subject during performance of the at least one activity, the instances of kinematic data being calculated based at least in part on updates made to the three-dimensional model depicting the subject over the period of time;
• and
• transmitting instructions that modify operations of one or more physical devices of the workspace based at least in part on the instances of kinematic data.

A non-transitory computer-readable storage medium comprising computer-executable instructions that, upon execution by one or more processors, cause one or more processors to perform operations comprising:

• obtaining instances of video capture data captured by one or more capture devices, each instance of video capture data depicting a two-dimensional representation of a subject during performance of at least one activity within a workspace;
• identifying a set of points of the subject during performance of the at least one activity based at least in part on providing the instances of video capture data to a machine-learning model as input, the machine-learning model being previously trained to identify points on a plurality of subjects in two-dimensional space from at least one image provided as input, the machine-learning model being trained utilizing a supervised learning algorithm and one or more historical images for which a corresponding set of points of a respective subject depicted in the historical image is known;
• electronically updating a three-dimensional model depicting the subject during the performance of the at least one activity, the three-dimensional model of the subject depicting performance of the at least one activity, the three-dimensional model being updated based at least in part on the set of points identified by the machine-learning model;
• calculating, over a period of time, instances of kinematic data related to the subject during performance of the at least one activity, the instances of kinematic data being calculated based at least in part on updates made to the three-dimensional model depicting the subject over the period of time;
• and
• transmitting instructions that modify operations of a physical device of the workspace based at least in part on the instances of kinematic data.