I'm using Simba for training classifiers on certain behaviors. When I generate the validation video after running the machine model, I realize that when body parts are occluded by an element of the arena or another mouse, they go all over the place. I was expecting an occluded body part to disappear, as we can decide of a Bp threshold in other parts of the pipeline (ROI analysis for example). You can see a picture below:
In this picture, a mouse is hidden in a nest located in the bottom right corner of the arena, as you can see, the points for the body parts did not disappear but go all over the place.
I am afraid this could mess with some classifications of behaviors and was therefore wondering if we could make the body part points disappear when they are occluded based on their detection probability.
I'm using Simba for training classifiers on certain behaviors. When I generate the validation video after running the machine model, I realize that when body parts are occluded by an element of the arena or another mouse, they go all over the place. I was expecting an occluded body part to disappear, as we can decide of a Bp threshold in other parts of the pipeline (ROI analysis for example). You can see a picture below:
In this picture, a mouse is hidden in a nest located in the bottom right corner of the arena, as you can see, the points for the body parts did not disappear but go all over the place.
I am afraid this could mess with some classifications of behaviors and was therefore wondering if we could make the body part points disappear when they are occluded based on their detection probability.
Thanks in advance.