Let’s say I have the frame coming from the camera. There is an image, that I would like to put on this frame so its perspective changes according to how the frame changes. I was able to achieve that using the object tracking with homography but here is the deal – overlaid image obstructs all the objects present on the frame. So for example if there is a person on that frame in the spot where this image should be located the person would not be visible. Is there any algorithm or a ML model that can help displaying the obstacles in the area where image will be? I tried using CV2 tools like findContours with dilation and stuff like this, but unfortunately it did not lead to a desired result.