AI Paper Atlas

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

MagicSeg: Open-World Segmentation Pretraining via Counterfactural Diffusion-Based Auto-Generation

Kaixin Cai, Pengzhen Ren, Jianhua Han, Yi Zhu, Hang Xu, Jianzhuang Liu, C X Liang

TPAMI 2026 · Image generation › Text-to-image diffusion · approach: Compositional text-to-image guidance · 0 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 ChatGPT-generated detailed prompts instead of simple 'a photo of class' prompt (ablation)MagicSeg (40k) · COCO valmIoU: 28.6 -> 36.0Improved
add counterfactual contrastive loss with MagicSeg prompts (ablation)MagicSeg (40k) · COCO valmIoU: 36.0 -> 37.2Improved
add counterfactual contrastive loss with simple prompts (ablation)MagicSeg (40k) · COCO valmIoU: 28.6 -> 27.5Hurt
add counterfactual contrastive loss on full dataset (ablation)MagicSeg (380k) · COCO valmIoU: 38.1 -> 40.2Improved
sample 100 categories per image instead of 1 (ablation)MagicSeg · COCO valmIoU: 7.7 -> 38.1Improved
sample 100 categories per image instead of all 1205 (ablation)MagicSeg · COCO valmIoU: 31.4 -> 38.1Improved
train on MagicSeg vs Grounded SAM baseline on LVISMagicSeg · LVIS valmIoU: 0.5 -> 4.9Improved
train on MagicSeg vs Grounded SAM baseline on PASCAL VOCMagicSeg · PASCAL VOC valmIoU: 61.8 -> 62.9Improved
scale training data from 20 classes (20k) to full 1205 classes (380k)MagicSeg · COCO valmIoU: 20.5 -> 40.2Improved
scale training data from 20 classes (20k) to full 1205 classes (380k)MagicSeg · PASCAL VOC valmIoU: 64.3 -> 62.9Hurt
use MagicSeg synthetic data vs Dataset Diffusion with DeepLabV3 R101DeepLabV3 R101 · PASCAL VOC valmIoU: 64.8 -> 65.3Improved
use MagicSeg synthetic data vs Dataset Diffusion with Mask2Former R50Mask2Former R50 · PASCAL VOC valmIoU: 60.2 -> 59.3Hurt

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