D-CRACKS: 1,000 Soil Desiccation Crack Images with Metadata

D-CRACKS compiles 1,000 images from laboratory tests of soil desiccation cracking collected from 45 studies. Chao-Sheng Tang published the dataset on figshare, last updated on April 10, 2026. Images are linked to metadata on testing conditions, soil properties, and analyzed crack properties.

Use Cases

• Developing predictive models for soil crack formation based on soil properties and environmental conditions.
• Validating numerical simulations of soil cracking based on extracted crack properties like area and length.
• Analyzing variability and trends in crack behavior across different soil types and testing conditions.
• Training machine learning models for crack detection and quantification from soil images.

Strengths

• Contains 1,000 images from 45 different studies, providing a broad sample.
• Each image is linked to detailed metadata including soil properties, testing conditions, and environmental conditions.
• Crack properties such as area, ratio, length, and width have been extracted and analyzed.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect bias inherent to the 45 selected literature studies.

Provenance

Source

figshare

Collection Method

Compiled from laboratory tests documented in 45 studies from the literature.

Freshness

Last updated 2026-04-10 07:04:48