An image generation apparatus comprises: a first-image obtaining unit adapted to obtain an image obtained by causing an image capturing unit to capture the grip unit controlled so as to place the target object in one predetermined orientation of a plurality of predetermined orientations with respect to the image capturing unit and the target object in the one predetermined orientation as a grip-state image; a second-image obtaining unit adapted to obtain, as a non-grip-state image corresponding to the one predetermined orientation, an image of the grip unit that does not grip the target object and is placed in a predetermined orientation coincident with the orientation controlled to place the target object in the one predetermined orientation; and an image generation unit adapted to generate a target object image including only the target object for the one predetermined orientation based on a difference between the grip-state image and the non-grip-state image.