Dataset Condensation Atlas

Method · Diffusion-based synthesis

D3S2

D3S2: Diffusion-Guided Dataset Distillation for Semantic Segmentation

Wenjie Zheng, Haoji Hu, Jiali Lu, Xingze Zou, Jing Wang

arXiv 2026 · first public 2026-05-24 · arXiv 2605.25022

paper ↗catalogued✓ abstract read

In one paragraph

Proposes D3S2 for dense-prediction dataset distillation in semantic segmentation: a class-balanced set of masks is selected by a greedy strategy that prioritizes underrepresented classes, then a pretrained layout-to-image diffusion model generates images conditioned on those masks, guided at sampling time by a segmentation-consistency loss and a class-wise feature-matching loss for pixel-level alignment; at a 1% compression rate with Mask2Former (Swin-S) it reports 24.99% mIoU on ADE20K and 35.49% on COCO-Stuff, improving over random selection by 9.34 and 5.70 points respectively.

Where it sits

Design choices

Prior / networks useddiffusion
Optimization regimetraining-free
Abstract (verbatim from arXiv)

Dataset distillation (DD) aims to compress large-scale datasets into compact synthetic sets while preserving training efficacy. However, existing studies mainly focus on image classification, leaving dense prediction tasks such as semantic segmentation largely underexplored. In this work, we identify three key challenges for segmentation DD: (i) long-tailed class imbalance, (ii) the need for strict pixel-wise alignment between images and dense labels, and (iii) the high computational cost of optimizing high-resolution data with complex models. To address these challenges, we propose D3S2, a Diffusion-guided Dataset Distillation framework for Semantic Segmentation. Our method adopts a two-stage design. In Class-Balanced Mask Selection, we construct a representative mask set via a greedy strategy that prioritizes underrepresented classes. In Diffusion-Guided Image Synthesis, we employ a pretrained layout-to-image diffusion model to generate images conditioned on the selected masks, naturally ensuring spatial alignment. To further enhance the training utility of synthesized data, we introduce guided diffusion sampling with two complementary objectives: a segmentation-consistency loss for pixel-level alignment, and a class-wise feature matching loss for aligning per-class feature statistics across layers. Extensive experiments demonstrate the superiority of D3S2. Notably, at an extremely compression rate of 1%, our method achieves 24.99% and 35.49% mIoU on ADE20K and COCO-Stuff with Mask2Former (Swin-S), outperforming random selection by 9.34% and 5.70%, respectively.

BibTeX (generated; prefer the venue's official entry)
@article{zheng2026diffusion,
  title   = {D3S2: Diffusion-Guided Dataset Distillation for Semantic Segmentation},
  author  = {Wenjie Zheng and Haoji Hu and Jiali Lu and Xingze Zou and Jing Wang},
  journal = {arXiv preprint arXiv:2605.25022},
  year    = {2026}
}

Nearby in Diffusion-based synthesis

2026-05

DMGD — DMGD: Train-Free Dataset Distillation with Semantic-Distribution Matching in Diffusion Models

Qichao Wang, Yunhong Lu, Hengyuan Cao et al. · CVPR 2026notablepaper ↗

2026-04

Learnability-guided diffusion — Learnability-Guided Diffusion for Dataset Distillation

Jeffrey A. Chan-Santiago, Mubarak Shah · CVPR 2026notablepaper ↗

2026-03

IMS3 — IMS3: Breaking Distributional Aggregation in Diffusion-Based Dataset Distillation

Chenru Wang, Yunyi Chen, Zijun Yang et al. · CVPR 2026notablepaper ↗

2026-03

EVLF — EVLF: Early Vision-Language Fusion for Generative Dataset Distillation

Wenqi Cai, Yawen Zou, Guang Li et al. · CVPR 2026notablepaper ↗code ↗

2026-02

ManifoldGD — ManifoldGD: Training-Free Hierarchical Manifold Guidance for Diffusion-Based Dataset Distillation

Ayush Roy, Wei-Yang Alex Lee, Rudrasis Chakraborty et al. · CVPR 2026notablepaper ↗code ↗