Research map / Detection and segmentation / Open-set and few-shot detection

MultiOrg: A Multi-rater Organoid-detection Dataset

NeurIPS 2024 · Detection and segmentation › Open-set and few-shot detection · approach: Remote sensing segmentation and change · 2 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Releases MultiOrg, a lung-organoid object-detection dataset of 411 brightfield well images (~63k boxes) with multiple expert label sets on the test split for label-uncertainty analysis, and benchmarks four detectors.

Problem

Organoid detection in microscopy lacks open datasets, and few biomedical detection datasets quantify annotation uncertainty via multiple raters.

Approach

A multi-rater organoid detection dataset in which the test images carry three label sets (test0 from two annotators at t0, and test1 A and B re-annotated at t1), enabling intra- and inter-rater uncertainty analysis, plus a benchmark and a napari plugin built on the best model.

Key results

• SSD achieves the best mAP@0.5 mean of 68.09% across the three label sets, while YOLOv3 has the highest mean F1-score of 0.62 and precision of 0.66.
• Faster R-CNN reaches the highest recall (mean 0.91) but very low precision (mean 0.19), giving an F1 of 0.31.
• SSD's best precision-recall curve is on test1_B, despite being trained on t0 labels, indicating the model agrees more with the later annotations.

Measured effects reported in the paper

Surprising findings

• The model trained on t0 labels agrees most with the t1_B label set, suggesting higher label noise in test0 was not captured by the model during training.
• Different detectors show very different precision-recall trade-offs at the standard 0.5 confidence threshold.

Limitations

• Organoid types were not annotated as separate classes, so the task is single-class detection.
• Multiple label sets cannot be used in the training loop; only test set has multiple labels.
• Only lung organoids from murine cells are included in the dataset; generalization to other organoid types is only preliminarily tested.

Experimental setup

Datasets: MultiOrg

Benchmarks: MultiOrg

Metrics: precision recall F1-score mAP@0.5IoU mAP@0.75IoU

Baselines: Faster R-CNN SSD YOLOv3 RTMDet

Builds on: Faster R-CNN SSD YOLOv3 RTMDet mmdetection napari-organoid-counter

Closest papers by meaning

• MorphoHELM: A Comprehensive Benchmark for Evaluating Representations for Microscopy-Based Morphology Assays (NeurIPS 2026 · similarity 0.86)
• PathView-Bench: Can Multimodal Large Language Models Achieve Fine-grained Multiscale Understanding of Pathology Images? (NeurIPS 2026 · similarity 0.84)
• STimage-1K4M: A histopathology image-gene expression dataset for spatial transcriptomics (NeurIPS 2024 · similarity 0.84)
• Semantic and Visual Crop-Guided Diffusion Models for Heterogeneous Tissue Synthesis in Histopathology (NeurIPS 2025 · similarity 0.84)
• BIOMEDICA: An Open Biomedical Image-Caption Archive, Dataset, and Vision-Language Models Derived from Scientific Literature (CVPR 2025 · similarity 0.84)
• Arctique: An artificial histopathological dataset unifying realism and controllability for uncertainty quantification (NeurIPS 2024 · similarity 0.83)