3D conversion relies on several cues. If the scene is cgi, it will have x,y and z coordinates already builtin where x and y are horizontal and vertical and z is depth, so the z coordinate will be used to determine the 3D aspect. Other cues are movement left and right. Infinity (objects farthest away) will be static (not move). As objects get closer to the camera they show more movement. The z coordinate is then extracted from that object displacement. Another cue is objects closest to you are usually closer from the ground and recede up to eye level, where eye level is inifinity. It takes a series of frames do determine depth, then each frame is extrapolated with the depth and extra pixels added for roundness. (From my years of experience running a VR center).