How output layer 'Sigma' values are reset after each sample in yolo_kp sigma delta network

[gwgknudayanga]

Hi,
In the yolo-kp sigma-delta network, each sample contain a set of images. while processing these set of images for a given sample, a sigma value is accumulated. So after processing these set of images for that particular sample the sigma value of the output layer needs to reset facilitating for the next sample.(because if sigma continously got accumulated it will overflow and then always gives a wrong value.). So how this resetting of sigma value is done after processing a given sample?

Thanks and Rgds,
Udayanga