{"success":true,"data":{"author":"Chákṣu IMAGE authors; mirrored by MedOtter","bucketAccessType":"public","bucketName":"MedOtter/Chaksu","bucketPrefix":"","category":"Health & Biomedical","cloudProvider":"huggingface","createdAt":"2026-11-17T19:29:33.433Z","description":"Chákṣu IMAGE is a glaucoma-specific color fundus image database. This mirror carries 1,345 fundus photographs (train 1,009, test 336) from three cameras: Remidio Fundus-on-Phone (2448x3264, 1,074 images), Bosch (1920x1440, 145) and Forus 3Nethra (2048x1536, 126). Five ophthalmologists annotated the optic disc and optic cup on every image; the STAPLE fusion is the primary label (mask with 0 background, 1 neuroretinal rim, 2 optic cup, plus binary disc and cup masks), with majority and median fusions as secondary columns. Each image also carries a glaucoma decision (normal or glaucoma suspect) by majority vote and the five per-expert decisions, plus disc, cup and rim areas and cup-to-disc measurements. Parquet files under data/ hold images at native resolution with unmodified bytes; about 2 GB (measured 2026-11-17). Use for optic disc and cup segmentation, cup-to-disc ratio estimation and glaucoma screening across devices. Mirrored on Hugging Face by MedOtter from the authors' release, with a card documenting provenance and any normalization. The authors' figshare record 20123135 is CC BY 4.0; cite the Chákṣu IMAGE article in Scientific Data (2023), doi 10.1038/s41597-023-01943-4.","displayTitle":"Chákṣu IMAGE: 1,345 glaucoma fundus photographs with optic disc and cup segmentations from five experts","doi":"10.6084/m9.figshare.20123135","huggingFaceUrl":"https://huggingface.co/datasets/MedOtter/Chaksu","id":"f9fe7432-5be1-57d0-9f1f-c376e4f28950","license":"CC-BY-4.0","licenseEvidenceUrl":"https://api.figshare.com/v2/articles/20123135","objectCountEstimate":12,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1038/s41597-023-01943-4","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/Chaksu\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"MedOtter/Chaksu\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=MedOtter, license:cc-by-4.0, size_categories:1K<n<10K, format:parquet, downloads=204"],"format":"mixed","generatedAt":"2026-11-17T19:40:29.844Z","layout":{"example":"data/test-00000-of-00002.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/test-00000-of-00002.parquet","truncatedAt":512}],"sampleObjects":["Readme_Chaksu_figshare.pdf","data/test-00000-of-00002.parquet","data/test-00001-of-00002.parquet","data/train-00000-of-00007.parquet","data/train-00001-of-00007.parquet"],"sidecars":[{"bytes":"---\nlicense: cc-by-4.0\ntask_categories:\n- image-segmentation\ntags:\n- medical\n- fundus\n- retina\n- glaucoma\n- optic-disc\n- optic-cup\nsize_categories:\n- 1K<n<10K\n---\n\n# Chaksu\n\nOptic **disc** and optic **cup** segmentation in colour fundus photographs, for\nglaucoma assessment. Mirror of the Chákṣu database for the MedOtter benchmark suite.\n\n| | |\n|---|---|\n| Modality | Colour fundus photography (2D RGB) |\n| Task | Optic disc + optic cup segmentation (also carries a glaucoma decision label) |\n| Images | **1,345** — train 1,009 / test 336 |\n| Cameras | Remidio FoP 2448×3264 (1,074) · Bosch 1920×1440 (145) · Forus 3Nethra 2048×1536 (126) |\n| Ground truth | **STAPLE** fusion of 5 expert annotators |\n| License | CC BY 4.0 |\n\n## Ground truth\n\nFive ophthalmologists annotated every image. The source release ships four fusions\nof those annotations; the paper designates **STAPLE** as the gold standard, having\nthe highest agreement with the individual experts:\n\n| Fusion | OD Dice | Role here |\n|---|---|---|\n| **STAPLE** | **0.9764** | **primary** → `mask`, `disc_mask`, `cup_mask` |\n| Majority (≥3 of 5) | 0.9732 | secondary → `*_mask_majority` |\n| Median | 0.9706 | secondary → `*_mask_median` |\n| Mean (all 5 agree) | 0.9503 | not mirrored (lowest agreement) |\n\n`mask` is a disjoint label map at native image resolution:\n\n| value | class |\n|---|---|\n| 0 | background |\n| 1 | neuroretinal rim (disc minus cup) |\n| 2 | optic cup |\n\nso `disc = mask > 0` and `cup = mask == 2`.\n\nSTAPLE masks are stored anti-aliased upstream; they are binarised at >127. In a\nsmall number of images the cup exceeded the disc by a handful of edge pixels, so\n`disc_mask` is stored as *disc ∪ cup* to guarantee cup ⊆ disc.\n\n## Columns\n\n| column | notes |\n|---|---|\n| `image` | original fundus photograph, native resolution, unmodified bytes |\n| `mask` | disjoint 0/1/2 label map (STAPLE) |\n| `disc_mask`, `cup_mask` | binary 0/255 (STAPLE) |\n| `disc_mask_majority`, `cup_mask_majority` | binary 0/255 (Majority fusion) |\n| `disc_mask_median`, `cup_mask_median` | binary 0/255 (Median fusion) |\n| `image_id`, `camera`, `split`, `height`, `width` | identifiers / geometry |\n| `glaucoma_decision` | majority vote of 5 experts: `NORMAL` / `GLAUCOMA SUSPECT` |\n| `decision_expert_1..5` | per-expert decision (whitespace/spelling normalised) |\n| `disc_area`, `cup_area`, `rim_area`, `cup_height`, `cup_width`, `disc_height`, `disc_width` | pixel measurements from the source CSVs |\n| `acdr`, `vcdr`, `hcdr` | area / vertical / horizontal cup-to-disc ratio |\n\n## Scope of this mirror\n\nThis is a **subset** of the full figshare release, restricted to what a\nsegmentation benchmark needs. Present: tier 1.0 (original images), tier 5.0\nSTAPLE/Majority/Median fusions, tier 6.0 metrics and decisions. **Not mirrored:**\ntier 2.0 (expert contour markups), tier 3.0 (per-expert binary masks — note\nExpert 3 has no Bosch folder), tier 4.0 (OD+OC fusion images), and the Mean and\nOverlay variants of tier 5.0. Those total ~200 GB of mostly uncompressed TIFF and\nare available from the original figshare deposit.\n\n## Caveats\n\n- **Patient overlap across the official split.** Only the Bosch subset carries\n  patient IDs, and at least one patient straddles the boundary (`P10_Image1` in\n  train, `P10_Image2` in test). Forus and Remidio splits are sequential cuts by\n  capture order, not patient-aware, so fellow-eye pairs may also straddle it.\n  The official split is preserved as published.\n- Source tier-6 CSVs are partly duplicated across the train and test folders\n  (the Bosch CSV is dataset-wide in both); they are merged by `(camera, image_id)`.\n- Expert decision strings upstream include a misspelling (`GLAUCOMA  SUSUPECT`);\n  these are normalised to `GLAUCOMA SUSPECT`.\n- No overlap with REFUGE, DRISHTI-GS, RIM-ONE, ORIGA, PAPILA, IDRiD, DRIVE or\n  RAVIR — Chaksu was newly acquired at KMC/MIT Manipal.\n\n## Citation\n\n```bibtex\n@article{kumar2023chaksu,\n  title   = {Chaksu: A glaucoma specific fundus image database},\n  author  = {Kumar, J. R. Harish and Seelamantula, Chandra Sekhar and\n         ","encoding":"utf8","key":"README.md","name":"README.md","size":4450,"truncated":true}],"sourceObjectKey":"data/test-00000-of-00002.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"MedOtter/Chaksu\")\nprint(ds)","starterSamplePath":"data/test-00000-of-00002.parquet","totalObjects":12},"summary":"Chákṣu IMAGE is a glaucoma-specific color fundus image database. This mirror carries 1,345 fundus photographs (train 1,009, test 336) from three cameras: Remidio Fundus-on-Phone…","title":"Chákṣu IMAGE: 1,345 glaucoma fundus photographs with optic disc and cup segmentations from five experts","updatedAt":"2026-11-17T23:00:22.589Z","visibility":"public","websiteUrl":"https://huggingface.co/datasets/MedOtter/Chaksu"}}