Sign in to view source links and access this dataset
Description
A civil-infrastructure visual inspection dataset for instance segmentation with 6 defect/condition categories: Algae, Crack, Net-Crack, Crack with Precipitation, Rust, and Spalling. Each sample is either a full-resolution inspection image or a 1024×1024 tile derived from one. The dataset was created by ibm-research and was last updated on Hugging Face in May 2026.
Use Cases
Train instance segmentation models to identify cracks in infrastructure based on the described defect categories.
Develop automated visual inspection systems for civil assets based on the full-resolution and tiled image samples.
Benchmark model performance on detecting complex defects like 'Crack with Precipitation' or 'Net-Crack'.
Study the effectiveness of tiling strategies for large-scale infrastructure image analysis as mentioned in the description.
Strengths
Provides 6 distinct defect categories for detailed classification.
Includes both full-resolution images and standardized 1024×1024 tiles.
Tiled samples carry extra metadata fields (tile_row, tile_col, file_name_original) for traceability.
Limitations
Row count, column details, and file formats are unknown, which may limit suitability assessment.
Column-level documentation is absent; field semantics must be inferred after download.
Data may reflect geographic or temporal bias inherent to the source inspection images.
Provenance
Source
ibm-research via Hugging Face
Collection Method
Likely gathered from visual inspections of civil infrastructure.
Time Range
null
Freshness
Last updated 2026-05-07 08:32:02; freshness should be verified.
Geography
null
License is unknown; restrictions should be verified before use.