Research map / Image generation / Text-to-image diffusion
EM3M: An Electron Micrograph Dataset for Microstructural Segmentation and Generation
Nan Wang, Zhiyi Xia, Yiming Li, Shi Tang, Zoe Fan, Xi Fang, Haoyi Tao, ZHANG SIYUAN, Guolin Ke, Yanhui Hong
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add 30K generated images to real training data (ablation) | UniAIMS trained on 4K real · EM3M (full test set) | mAP@0.5: 0.800 -> 0.809 | Improved |
| add 30K generated images to real training data (ablation) | UniAIMS trained on 4K real · EM3M (full test set) | PQ@0.5: 0.704 -> 0.707 | Improved |
| increase real training data from 1K to 4K images (ablation) | UniAIMS · EM3M (full test set) | mAP@0.5: 0.767 -> 0.800 | Improved |
| increase generated training data from 5K to 30K images (ablation) | UniAIMS · EM3M (full test set) | mAP@0.5: 0.760 -> 0.791 | Improved |
| replace 3K real images by 30K generated images (ablation) | UniAIMS · EM3M (full test set) | mAP@0.5: 0.795 (3K real) vs 0.791 (30K generated) | No clear effect |
| pretrain on 30K generated images instead of SA1B (ablation) | UniAIMS · EM3M (full test set) | mAP@0.5 | Improved |
| finetune on external dataset after zero-shot | UniAIMS · Okunev (external) | mAP: 0.81 -> 0.84 | Improved |
| finetune on external dataset after zero-shot | UniAIMS · EMPS (external) | mAP: 0.89 -> 0.92 | Improved |
| finetune SDXL with LoRA on EM3M structured prompts | SDXL · Generated vs real EM | FID: 213.16 -> 34.03 | Improved |
| finetune SDXL with LoRA on EM3M structured prompts | SDXL · Generated vs real EM | Style similarity: 0.669 -> 0.958 | Improved |
| use flow-based UniAIMS (no size model, LSJ augmentation, modern backbone) instead of Cellpose | Cellpose · EM3M (sparse subset, <100 instances) | mAP@0.5: 0.650 -> 0.818 | Improved |
| use UniAIMS instead of Cellpose | Cellpose · EM3M (full test set) | mAP@0.5: 0.711 -> 0.800 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SDXL base — SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis (ICLR 2024) · better on 2 of 2 shared results
- CPP-Net — CPP-Net: Embracing Multi-Scale Feature Fusion into Deep Unfolding CP-PPA Network for Compressive Sensing (CVPR 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ATOMIC: A Domain-Specific Vision-Language Model for Transmission Electron Microscopy (ECCV 2026 · similarity 0.88)
- Learning Large-Factor EM Image Super-Resolution with Generative Priors (CVPR 2024 · similarity 0.84)
- Scalable Diffusion for Materials Generation (ICLR 2024 · similarity 0.84)
- Periodic Materials Generation using Text-Guided Joint Diffusion Model (ICLR 2025 · similarity 0.83)
- $\epsilon$-Seg: Sparsely Supervised Semantic Segmentation of Microscopy Data (NeurIPS 2025 · similarity 0.83)
- Co-PiLOT: Constrained Physics-Informed Latent Optimization for Target-Driven Inverse Design (NeurIPS 2026 · similarity 0.83)
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.
