| CPC G06T 7/11 (2017.01) [A01K 1/031 (2013.01); A01K 29/005 (2013.01); G06F 18/28 (2023.01); G06V 10/25 (2022.01); G06V 10/26 (2022.01); G06V 10/454 (2022.01); G06V 10/764 (2022.01); G06V 10/772 (2022.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01); G06V 20/00 (2022.01)] | 14 Claims |

|
1. A method of training a neural network architecture comprising:
receiving, at a computing system, a first plurality of frames depicting a mouse in a laboratory environment;
receiving, at the computing system, at least a mouse label and a background label on each of the first plurality of frames;
predicting, using the computing system, a segmentation corresponding to the mouse and an ellipse corresponding to the mouse in each of the first plurality of frames;
in response to predicting the segmentation corresponding to the mouse and the ellipse corresponding to the mouse in each of the first plurality of frames, receiving a plurality of additional labels on the first plurality of frames;
predicting, using the computing system, an updated segmentation corresponding to the mouse and an updated ellipse corresponding to the mouse in each of the first plurality of frames;
determining, using the computing system, that the predictions are within a predetermined error tolerance;
receiving, at the computing system, a direction of the ellipse in each of the first plurality of frames;
storing the first plurality of frames as an annotated data set; and
training the neural network architecture with the annotated data set and the first plurality of frames depicting the mouse in the laboratory environment to determine direction and location of at least one mouse in a second plurality of frames.
|