InstructDiffusion: A Generalist Modeling Interface for Vision Tasks
This work aims to create a generalist modeling interface for vision tasks, advancing towards artificial general intelligence in computer vision, though it appears incremental as it builds upon diffusion processes.
The paper tackles the problem of aligning diverse computer vision tasks with human instructions by proposing InstructDiffusion, a framework that casts tasks into an image-manipulating process in pixel space, and it outperforms prior methods on novel datasets.
We present InstructDiffusion, a unifying and generic framework for aligning computer vision tasks with human instructions. Unlike existing approaches that integrate prior knowledge and pre-define the output space (e.g., categories and coordinates) for each vision task, we cast diverse vision tasks into a human-intuitive image-manipulating process whose output space is a flexible and interactive pixel space. Concretely, the model is built upon the diffusion process and is trained to predict pixels according to user instructions, such as encircling the man's left shoulder in red or applying a blue mask to the left car. InstructDiffusion could handle a variety of vision tasks, including understanding tasks (such as segmentation and keypoint detection) and generative tasks (such as editing and enhancement). It even exhibits the ability to handle unseen tasks and outperforms prior methods on novel datasets. This represents a significant step towards a generalist modeling interface for vision tasks, advancing artificial general intelligence in the field of computer vision.