Research map / Detection and segmentation / CNN features
Tiling artifacts and trade-offs of feature normalization in the segmentation of large biological images
Elena Buglakova, Anwai Archit, Edoardo D'Imprima, Julia Mahamid, Constantin Pape, Anna Kreshuk
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 |
|---|---|---|---|
| use BatchRenorm instead of InstanceNorm (ablation) | 3D U-Net · organoids matrix | Dice (eval mode): 0.61 -> 0.85 | Improved |
| use BatchRenorm instead of InstanceNorm (ablation) | 3D U-Net · embryo foreground | Dice (eval mode): 0.67 -> 0.74 | Improved |
| use BatchRenorm instead of InstanceNorm (ablation) | 3D U-Net · organoids boundaries | Dice (eval mode): 0.75 -> 0.73 | Hurt |
| use BatchRenorm instead of BatchNorm (ablation) | 3D U-Net · organoids boundaries | Dice (eval mode): 0.49 -> 0.73 | Improved |
| use BatchRenorm instead of BatchNorm (ablation) | 3D U-Net · organoids | train/eval disparity: 0.48 -> 0 | Improved |
| use BatchRenorm instead of InstanceNorm (ablation) | 3D U-Net · organoids | tile mismatch: 0.11 -> no | Improved |
| use BatchRenorm instead of InstanceNorm (ablation) | 3D U-Net · embryo | tile mismatch: 0.16 -> no | Improved |
| use BatchRenorm instead of InstanceNorm (ablation) | UNETR · organoids matrix | Dice (eval mode): 0.10 -> 0.87 | Improved |
| use BatchRenorm instead of BatchNorm (ablation) | UNETR · organoids matrix | Dice (eval mode): 0.03 -> 0.87 | Improved |
| use BatchRenorm instead of InstanceNorm (ablation) | 3D U-Net transfer embryo->plants · plants foreground | Dice (eval mode): 0.50 -> 0.50 | No clear effect |
| use BatchRenorm instead of InstanceNorm (ablation) | 3D U-Net transfer plants->embryo · embryo foreground | Dice (eval mode): 0.54 -> 0.59 | Improved |
| remove feature normalization (Identity) instead of BatchRenorm (ablation) | 3D U-Net · plants foreground | Dice (eval mode): 0.89 -> 0.52 | Hurt |
Closest papers by meaning
- $\epsilon$-Seg: Sparsely Supervised Semantic Segmentation of Microscopy Data (NeurIPS 2025 · similarity 0.84)
- UniverSeg: Universal Medical Image Segmentation (ICCV 2023 · similarity 0.84)
- High-Performance Large-Scale Image Recognition Without Normalization (ICML 2021 · similarity 0.84)
- Combinatorial Energy Learning for Image Segmentation (NeurIPS 2016 · similarity 0.84)
- Uncertainty-aware Fine-tuning of Segmentation Foundation Models (NeurIPS 2024 · similarity 0.83)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · 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.
