I have a set of images of dogs (around 80000 images). Some images contain full pictures of a dog and some are not and only half of the body is visible or partially visible (like only head and neck of a dog, or only a back leg ad half of the body). I want to divide images into two categories of full vs not full body, to only keep images that show full picture of a dog. Is there any NN model or approach that can differentiate between full and partial object?

I tried some object detection model, but they only detect the object and their bounding box is not useful as images are in different perspective and sometime an image shows a full dog but since it’s located far away, the box is small and conversely sometimes a half body of a dog has a bigger bounding box.