Published on:
18 January 2024
Primary Category:
Computer Vision and Pattern Recognition
Paper Authors:
Wouter Van Gansbeke,
Bert De Brabandere
Proposes latent diffusion for panoptic segmentation
Uses shallow autoencoder to compress masks
Learns conditional diffusion process
Enables mask completion capability
Achieves promising panoptic segmentation results
Simple generative model for panoptic segmentation
This paper proposes a simple generative approach using latent diffusion models for panoptic segmentation. It trains a shallow autoencoder to compress segmentation masks into a latent space, then learns an image-conditioned diffusion process to generate masks. Key advantages are simplicity, generality, and enabling mask completion applications.
Binary noise for anomaly detection
Diffusion Models as Masked Autoencoders: A New Approach for Learning Visual Representations
Diffusion models for medical image segmentation
Synthesizing satellite images and labels with diffusion models
Diffusion models learn visual representations
Sequential image generation by diffusing pixel groups
No comments yet, be the first to start the conversation...
Sign up to comment on this paper