ConcreteCARB

The ConcreteCARB dataset contains 903 high-resolution images of concrete samples tested for carbonation damage using the phenolphthalein indicator method under laboratory conditions.
The dataset is organized into two categories:
Carbonated Samples (431 images)
No Carbonation Presence (472 images)
Each image captures the fractured surface of a concrete prism after application of phenolphthalein, which visually distinguishes carbonated (colorless) and non-carbonated (pink) zones. The samples were prepared using various concrete mix designs, including different water–cement ratios and additives such as industrial silica waste and Opuntia ficus-indica extract. After 180 days of natural atmospheric exposure, the specimens were split using a manual chisel and hammer method and photographed using a Samsung SM-S901U1 smartphone (4000×3000 resolution, ambient lighting, no flash).
All images were manually reviewed and labeled by expert engineers. This dataset supports research in machine learning, deep learning, and computer vision applications for structural health monitoring, particularly in tasks related to carbonation detection, classification, and segmentation.
A comprehensive metadata file (Metadata.csv) accompanies the image dataset. This file links each image filename to its corresponding Specimen ID, Mixture Type (1, 2, or 3), Water-Cement Ratio, Additives (Industrial silica waste or Opuntia ficus-indica), Test age (180 days), and Exposure Condition. Technical photography details, including the camera model (Samsung Galaxy S22) and approximate imaging distance, are also recorded for each entry.