{"success":true,"data":{"author":"Wei Zheng et al. (HF copy by MedOtter)","bucketAccessType":"public","bucketName":"MedOtter/NIS3D","bucketPrefix":"","category":"Life Sciences","cloudProvider":"huggingface","createdAt":"2026-11-17T14:12:21.859Z","description":"NIS3D (NeurIPS 2023 Datasets and Benchmarks, Zheng et al.) is a dense 3D nuclei instance segmentation benchmark for fluorescence microscopy of embryonic tissue. Six large volumes (two each of zebrafish, Drosophila and Mus musculus embryos) carry more than 22,000 manually annotated nuclei vetted by three independent annotators. Each sample folder under NIS3D/ has data.tif (raw 3D fluorescence volume), GroundTruth.tif (consensus instance label map), ConfidenceScore.tif (per-cell reliability from inter-annotator agreement; voxels marked 1 have indeterminate boundaries and are ignored in evaluation), a visualization stack and Info.txt with microscope and resolution details. The repository also includes the suggested train/test splitting and a converted data/ copy; about 9 GB. Multi-page TIFFs open in Fiji or tifffile. This is a Hugging Face copy of the authors' Zenodo record (10.5281/zenodo.11456029), which is released under CC BY 4.0; cite Zheng et al. 2023 and github.com/yu-lab-vt/NIS3D.","displayTitle":"NIS3D: densely annotated 3D nuclei segmentation benchmark in zebrafish, Drosophila and mouse embryos","doi":"10.5281/zenodo.11456029","huggingFaceUrl":"https://huggingface.co/datasets/MedOtter/NIS3D","id":"80b51407-7796-52ed-9ed8-846a5b5c4b4e","license":"CC-BY-4.0","licenseEvidenceUrl":"https://zenodo.org/records/11456029","objectCountEstimate":102,"probeStatus":"ok","referenceUrl":"https://github.com/yu-lab-vt/NIS3D","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/NIS3D\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"MedOtter/NIS3D\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=MedOtter, license:cc-by-4.0, size_categories:n<1K, format:parquet, downloads=1241"],"format":"mixed","generatedAt":"2026-11-17T14:23:13.920Z","layout":{"example":"data/train-00000-of-00001.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/train-00000-of-00001.parquet","truncatedAt":512}],"sampleObjects":["CONVERSION.md","NIS3D/Drosophila_1/ConfidenceScore.tif","NIS3D/Drosophila_1/GroundTruth.nii.gz","NIS3D/Drosophila_1/GroundTruth.tif","NIS3D/Drosophila_1/Info.txt"],"sidecars":[{"bytes":"---\nlicense: cc-by-4.0\ntask_categories:\n- image-segmentation\nmodality:\n- microscopy\nlanguage: []\ntags:\n- microscopy\n- fluorescence\n- 3D\n- nuclei-segmentation\n- instance-segmentation\n- embryonic-imaging\n- NeurIPS-2023\npretty_name: NIS3D\nsize_categories:\n- n<1K\ndataset_info:\n  features:\n  - name: sample_id\n    dtype: string\n  - name: subset\n    dtype: string\n  - name: num_slices\n    dtype: int32\n  - name: height\n    dtype: int32\n  - name: width\n    dtype: int32\n  - name: cells_in_middle_slice\n    dtype: int32\n  - name: data_middle_slice\n    dtype: image\n  - name: gt_middle_slice\n    dtype: image\n  - name: confidence_middle_slice\n    dtype: image\n  - name: overlay_middle_slice\n    dtype: image\n  splits:\n  - name: train\n    num_bytes: 41481448\n    num_examples: 24\n  download_size: 41487097\n  dataset_size: 41481448\nconfigs:\n- config_name: default\n  data_files:\n  - split: train\n    path: data/train-*\n---\n\n# NIS3D — A Completely Annotated Benchmark for Dense 3D Nuclei Image Segmentation\n\nNIS3D is a dense 3D nuclei segmentation benchmark for fluorescence microscopy of embryonic tissue. It provides 6 large-volume 3D images (Zebrafish, Drosophila, Mus Musculus) with **22,000+ manually annotated cells** vetted by three independent annotators, accompanied by per-cell confidence scores derived from inter-annotator agreement.\n\nPublished at **NeurIPS 2023 (Datasets and Benchmarks Track)**.\n\n## Dataset Summary\n\n| Field | Details |\n|---|---|\n| Modality | 3D fluorescence microscopy (volumetric TIFF stacks) |\n| Target | Cell nuclei (instance segmentation) |\n| Species | Zebrafish, Drosophila, Mus Musculus (mouse) |\n| Samples | 6 (Zebrafish_1, Zebrafish_2, Drosophila_1, Drosophila_2, MusMusculus_1, MusMusculus_2) |\n| Annotated cells | 22,000+ |\n| File format | Multi-page TIFF (`.tif`), opens with Fiji/ImageJ or `tifffile` |\n| License | CC-BY-4.0 |\n| Source | Zenodo: https://zenodo.org/records/11456029 (DOI 10.5281/zenodo.11456029) |\n| GitHub | https://github.com/yu-lab-vt/NIS3D |\n\n## Per-Sample Files\n\nEach sample folder under `NIS3D/` contains:\n\n| File | Role |\n|---|---|\n| `data.tif` | Raw 3D fluorescence volume |\n| `GroundTruth.tif` (or `gt.tif` for MusMusculus_2) | **Recommended ground-truth instance label map** (different integers = different cells, 0 = background) |\n| `ConfidenceScore.tif` (or `scoreOfConfidence.tif` for MusMusculus_2) | Per-cell reliability map: 0 = background, 1 = \"undefined mask\" (boundaries indeterminate, ignored at evaluation), 2 = 1/3 annotator agreement, 3 = 2/3, 4 = 3/3 (full consensus) |\n| `visulize.tif` | Composite visualization (raw + GT + confidence) — display only |\n| `Info.txt` | Acquisition details (microscope, resolution, pre-processing, etc.) |\n\nThe single `GroundTruth.tif` is the consensus across three annotators; per-annotator disagreement is encoded into `ConfidenceScore.tif` rather than rival GT files. **At evaluation, predictions falling on confidence-1 voxels should be ignored** (neither TP nor FP) because boundaries there are too weak for any annotator to delineate.\n\n## Splits\n\nThe Zenodo archive ships **two suggestive split schemes** under `suggestive splitting/`, in addition to the canonical full-size 6 samples in `NIS3D/`:\n\n| Split scheme | Description |\n|---|---|\n| `cross-image/` | Leave-one-per-species: train on *_2, test on *_1 (full-size volumes) |\n| `in-image/` | Each of the 6 samples is cropped into disjoint train/test halves |\n\nBoth schemes are advisory — researchers may use the canonical `NIS3D/` folder and define their own splits.\n\n## Per-Sample Stats\n\n| Sample | data.tif size | GroundTruth.tif size | Cells (approx.) |\n|---|---|---|---|\n| Zebrafish_1 | 1.25 GB | 68 MB | (high density) |\n| Zebrafish_2 | 45 MB | 6.5 MB | |\n| Drosophila_1 | 382 MB | 28 MB | |\n| Drosophila_2 | 386 MB | 17 MB | |\n| MusMusculus_1 | 554 MB | 45 MB | |\n| MusMusculus_2 | 273 MB | 47 MB | |\n\nPer-sample acquisition details (microscope, resolution, pre-processing) are in each `Info.txt`.\n\n## Citation\n\n```bibtex\n@inproceedings{zheng2023nis3d,\n  title     = {NIS3D: A Completely Annotated Benchmark for Dense 3D","encoding":"utf8","key":"README.md","name":"README.md","size":4856,"truncated":true}],"sourceObjectKey":"data/train-00000-of-00001.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/NIS3D\")\nprint(ds)","starterSamplePath":"NIS3D/Drosophila_1/ConfidenceScore.tif","totalObjects":102},"summary":"NIS3D (NeurIPS 2023 Datasets and Benchmarks, Zheng et al.) is a dense 3D nuclei instance segmentation benchmark for fluorescence microscopy of embryonic tissue.","title":"NIS3D: densely annotated 3D nuclei segmentation benchmark in zebrafish, Drosophila and mouse embryos","updatedAt":"2026-11-17T20:08:19.902Z","visibility":"public","websiteUrl":"https://huggingface.co/datasets/MedOtter/NIS3D"}}