Crack Images with Annotated Masks for Digital Image Correlation Analysis

Crack images and corresponding annotated ground truth masks used to train, validate, and test a deep convolutional neural network for crack pixel detection. The data was created by Amir Hossein Rezaie of École Polytechnique Fédérale de Lausanne and is associated with publications from 2020 and 2021. The dataset is hosted on the paperswithcode platform.

Use Cases

• Training crack segmentation models based on annotated ground truth masks.
• Validating digital image correlation (DIC) measurements against deep learning predictions.
• Benchmarking computer vision algorithms for defect detection in civil engineering materials.

Strengths

• Includes corresponding annotated ground truth masks for supervised learning.
• Data is associated with peer-reviewed publications from 2020 and 2021.
• Source code to reproduce the results is shared, supporting reproducibility.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

École Polytechnique Fédérale de Lausanne

Collection Method

Images taken as input for the digital image correlation (DIC) method, with annotations created for model training.