{
  "datasets": [
    {
      "name": "CrackVision12K",
      "short": "CV12K",
      "publisher": "Joo et al., 2025",
      "publication": "UCL Research Data Repository, 2024",
      "homepage": "https://rdr.ucl.ac.uk/articles/dataset/CrackVision12K/26946472",
      "domain": "Road pavement",
      "images": "12,000 (9,600 train / 1,200 val / 1,200 test)",
      "mask_convention": "crack = 255 (white) on black background",
      "license": "see source",
      "note": "Visit the source link for the canonical download and license terms."
    },
    {
      "name": "OmniCrack30K",
      "short": "OC30K",
      "publisher": "Benz & Rodehorst, 2024",
      "publication": "CVPRW 2024",
      "homepage": "https://github.com/ben-z-original/omnicrack30k",
      "domain": "Multi-domain (road, masonry, concrete, asphalt)",
      "images": "~30,000 across train / val / test splits (a subset of files use a 1-pixel-skeleton annotation style and are typically filtered)",
      "mask_convention": "crack = 0 (black) on white background — inverse of most other datasets",
      "license": "GPL-3.0",
      "note": "Redistribution must inherit GPL-3.0 and include attribution. Visit the source."
    },
    {
      "name": "TACK Tunnel Data (TTD)",
      "short": "TTD",
      "publisher": "Sjölander et al., 2025",
      "publication": "arXiv:2512.14477",
      "homepage": "https://huggingface.co/datasets/CrackedCity/tack-tunnel-data",
      "domain": "Tunnel linings (cracks, water, leaching)",
      "images": "3,774 images (785 with cracks, 197 water, 316 leaching)",
      "mask_convention": "Multi-class single-band PNG: 0=background, 40=crack, 160=water, 200=leaching",
      "license": "see source",
      "note": "Visit the source for the canonical download and license terms."
    },
    {
      "name": "CrackStructures (and related close-up sets)",
      "short": "CrackStructures",
      "publisher": "Benz et al., 2022",
      "publication": "n/a",
      "homepage": "https://github.com/ben-z-original/crackstructures",
      "domain": "Close-up crack imagery on structural surfaces",
      "images": "Hundreds of high-resolution structural close-ups",
      "mask_convention": "Binary masks; check source for exact convention",
      "license": "see source",
      "note": "Often used as a held-out, out-of-distribution evaluation set."
    },
    {
      "name": "CRACK500",
      "short": "CRACK500",
      "publisher": "Yang et al., 2019",
      "publication": "Feature Pyramid and Hierarchical Boosting Network for Pavement Crack Detection, TITS",
      "homepage": "https://github.com/fyangneil/pavement-crack-detection",
      "domain": "Smartphone pavement",
      "images": "500 source images, split into ~3,368 patches",
      "mask_convention": "Binary, crack = 255",
      "license": "see source",
      "note": "Widely cited baseline; visit the source for the canonical download."
    },
    {
      "name": "DeepCrack (Liu et al.)",
      "short": "DeepCrack",
      "publisher": "Liu et al., 2019",
      "publication": "Neurocomputing",
      "homepage": "https://github.com/yhlleo/DeepCrack",
      "domain": "Pavement and concrete",
      "images": "537 (300 train / 237 test)",
      "mask_convention": "Binary masks, crack = 255",
      "license": "see source",
      "note": "Small but high-quality benchmark."
    },
    {
      "name": "SegCODEBRIM",
      "short": "SegCODEBRIM",
      "publisher": "Mundt et al., 2019 (CODEBRIM) + segmentation re-annotation",
      "publication": "CVPR 2019",
      "homepage": "https://zenodo.org/record/2620293",
      "domain": "Reinforced concrete bridge damage",
      "images": "Hundreds of close-up concrete shots with multi-label damage",
      "mask_convention": "Standard binary (crack = 255)",
      "license": "see source",
      "note": "Original CODEBRIM is classification; the seg re-annotation is community-driven."
    }
  ]
}