I have created four 3D-CNN models and each was trained on a different (but related) set of images, such that each set has images of a different perspective of the same objec