Research map / Label-efficient and robust learning / Self-supervised and contrastive
MAESTER: Masked Autoencoder Guided Segmentation at Pixel Resolution for Accurate, Self-Supervised Subcellular Structure Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use MAESTER instead of Han et al. | Self-supervised segmentation on betaSeg · betaSeg test cell stack, K=4–10 | DSC: 0.625 -> 0.696; 0.659 -> 0.773; 0.647 -> 0.787; 0.643 -> 0.793; 0.560 -> 0.723; 0.578 -> 0.739; 0.567 -> 0.708 | Helped |
| Train with full ground-truth labels using Segmenter instead of self-supervised MAESTER | betaSeg segmentation · betaSeg test cell stack, nucleus, granules, mitochondria, unrecognized | DSC: 0.950 -> 0.990; 0.556 -> 0.860; 0.786 -> 0.896; 0.844 -> 0.912 | Mixed |
| Train with full ground-truth labels using Vanilla ViT instead of self-supervised MAESTER | betaSeg segmentation · betaSeg test cell stack, nucleus, granules, mitochondria, unrecognized | DSC: 0.950 -> 0.981; 0.556 -> 0.774; 0.786 -> 0.868; 0.844 -> 0.907 | Mixed |
| Increase FOV from 8×8 patches to 16×16 patches while keeping patch size constant (ablation) | MAESTER · betaSeg test cell stack, K=6 | DSC: 0.621 -> 0.787 | Helped |
| Increase masking ratio from 0.25 to 0.5 (ablation) | MAESTER · betaSeg test cell stack, K=6 | DSC: 0.685 -> 0.787 | Helped |
| Increase masking ratio from 0.5 to 0.85 (ablation) | MAESTER · betaSeg test cell stack, K=6 | DSC: 0.787 -> 0.721 | Hurt |
| Keep 6×6 rather than 4×4 tokens per FOV during inference (ablation) | MAESTER · betaSeg test cell stack, K=6 | DSC: 0.787 -> 0.753 | Hurt |
| Keep 8×8 rather than 4×4 tokens per FOV during inference (ablation) | MAESTER · betaSeg test cell stack, K=6 | DSC: 0.787 -> 0.750 | Hurt |
| Increase patch size from 3×3 px to 5×5 px (ablation) | MAESTER · betaSeg test cell stack, K=6 | DSC: 0.371 -> 0.787 | Helped |
| Increase patch size from 5×5 px to 7×7 px (ablation) | MAESTER · betaSeg test cell stack, K=6 | DSC: 0.787 -> 0.731 | Hurt |
| Increase patch size from 5×5 px to 9×9 px (ablation) | MAESTER · betaSeg test cell stack, K=6 | DSC: 0.787 -> 0.688 | Hurt |
| Add positional embedding (ablation) | MAESTER · betaSeg test cell stack, K=6 | DSC: 0.647 -> 0.787 | Helped |
Closest papers by meaning
- Masked Autoencoders for Microscopy are Scalable Learners of Cellular Biology (CVPR 2024 · similarity 0.88)
- SCCS: Deep Neural Spectral Clustering for Self-Supervised Subcellular Structure Segmentation (AAAI 2025 · similarity 0.87)
- Structure-Guided Masked Autoencoders for Ultra-High Resolution Scientific Image Understanding (NeurIPS 2026 · similarity 0.86)
- $\epsilon$-Seg: Sparsely Supervised Semantic Segmentation of Microscopy Data (NeurIPS 2025 · similarity 0.86)
- Beyond ViT Tokens: Masked-Diffusion Pretrained Convolutional Pathology Foundation Model for Cell-Level Dense Prediction (NeurIPS 2026 · similarity 0.85)
- ViTally Consistent: Scaling Biological Representation Learning for Cell Microscopy (ICML 2025 · similarity 0.85)
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.