AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Open-Vocabulary Panoptic Segmentation With Text-to-Image Diffusion Models

Jiarui Xu, Sifei Liu, Arash Vahdat, Wonmin Byeon, Xiaolong Wang, Shalini De Mello

CVPR 2023 · Image generation › Text-to-image diffusion · approach: Compositional text-to-image guidance · 354 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use diffusion features rather than CLIP(H) visual features (ablation)ODISE · ADE20K panoptic segmentation (512² ablation)PQ: 21.2 -> 23.3Improved
Generate an implicit caption instead of using an empty caption (ablation)ODISE · ADE20K panoptic segmentationPQ: 21.8 -> 23.3Improved
Use an implicit captioner instead of a heuristic captioner (ablation)ODISE · ADE20K panoptic segmentationPQ: 22.2 -> 23.3Improved
Use an implicit captioner instead of BLIP (ablation)ODISE · ADE20K panoptic segmentationPQ: 22.3 -> 23.3Improved
Use an implicit captioner instead of BLIP (ablation)ODISE · COCO panoptic segmentationPQ: 44.1 -> 44.2No clear effect
Increase the diffusion time step from 0 to 100 (ablation)ODISE · ADE20K panoptic segmentationPQ: 23.3 -> 22.8Hurt
Increase the diffusion time step from 0 to 200 (ablation)ODISE · ADE20K panoptic segmentationPQ: 23.3 -> 21.5Hurt
Increase the diffusion time step from 0 to 500 (ablation)ODISE · ADE20K panoptic segmentationPQ: 23.3 -> 20.9Hurt
Concatenate features from time steps 0, 100, and 200 rather than using step 0 alone (ablation)ODISE · ADE20K panoptic segmentationPQ: 23.3 -> 23.1No clear effect
Learn the diffusion time step rather than fixing it at 0 (ablation)ODISE · ADE20K panoptic segmentationPQ: 23.3 -> 22.8Hurt
Add discriminative-model predictions to diffusion-model predictions (ablation)ODISE · ADE20K panoptic segmentationPQ: 20.1 -> 23.3Improved
Add discriminative-model predictions to diffusion-model predictions (ablation)ODISE · COCO panoptic segmentationPQ: 42.3 -> 44.2Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy