Why do we only focus on 3D images here? How about 2D image generation? Or, 2D = 3D - projection?
air-wreck
Thinking of image generation as a series of stages in a pipeline is conceptually clean, but are there any performance benefits from taking "bigger steps" that might be messier?
Why do we only focus on 3D images here? How about 2D image generation? Or, 2D = 3D - projection?
Thinking of image generation as a series of stages in a pipeline is conceptually clean, but are there any performance benefits from taking "bigger steps" that might be messier?