Research map / Detection and segmentation / Open-set and few-shot detection
A Probabilistic U-Net for Segmentation of Ambiguous Images
Simon Kohl, Bernardino Romera-Paredes, Clemens Meyer, Jeffrey De Fauw, Joseph R. Ledsam, Klaus Maier-Hein, S. M. Ali Eslami, Danilo Jimenez Rezende, Olaf Ronneberger
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 a context-conditioned probabilistic latent prior and inject sampled latents late in the U-Net | Probabilistic U-Net · LIDC-IDRI and Cityscapes | squared GED | Improved |
| Use Probabilistic U-Net samples to distinguish ambiguous from unambiguous lesions | LIDC-IDRI · LIDC-IDRI test set | ambiguous-lesion discrimination accuracy: 0.699 -> 0.736 | Improved |
| Fix the prior to a standard Gaussian instead of conditioning it on the image (ablation) | Probabilistic U-Net · LIDC-IDRI and Cityscapes | squared GED | Hurt |
| Fix the prior and remove the input image from the posterior (ablation) | Probabilistic U-Net · LIDC-IDRI and Cityscapes | squared GED | Hurt |
| Inject latent features at the beginning rather than after the U-Net decoder (ablation) | Probabilistic U-Net · LIDC-IDRI and Cityscapes | squared GED | Hurt |
| Inject latent features at the beginning rather than after the U-Net decoder (ablation) | Probabilistic U-Net · multiple samples | squared GED | Hurt |
| Increase the number of samples drawn from the model | Probabilistic U-Net · LIDC-IDRI test set | squared GED: 0.811 -> 0.388 -> 0.321 -> 0.287 | Improved |
| Increase the number of samples drawn from the model | Probabilistic U-Net · Cityscapes validation set | squared GED: 0.874 -> 0.337 -> 0.248 -> 0.206 | Improved |
Closest papers by meaning
- U-KAN Makes Strong Backbone for Medical Image Segmentation and Generation (AAAI 2025 · similarity 0.86)
- Rethinking U-Net: Task-Adaptive Mixture of Skip Connections for Enhanced Medical Image Segmentation (AAAI 2025 · similarity 0.86)
- A Generative Approach at the Instance-Level for Image Segmentation Under Limited Training Data Conditions (Student Abstract) (AAAI 2025 · similarity 0.86)
- Generative Medical Segmentation (AAAI 2025 · similarity 0.86)
- Enhancing SAM with Efficient Prompting and Preference Optimization for Semi-supervised Medical Image Segmentation (CVPR 2025 · similarity 0.85)
- nnWNet: Rethinking the Use of Transformers in Biomedical Image Segmentation and Calling for a Unified Evaluation Benchmark (CVPR 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.
