DICEPTION: A Generalist Diffusion Model for Vision Perception

For non-human image inputs, the pose results may have issues. Same when perform semantic segmentation with categories that are not in COCO.

The results of semantic segmentation may be unstable because:

• We only trained on COCO, whose quality and quantity are insufficient to meet the requirements.

• Semantic segmentation is more complex than other tasks, as it requires accurately learning the relationship between semantics and objects.

However, we are still able to produce some high-quality semantic segmentation results, strongly demonstrating the potential of our approach.