Concrete Micro-Surface Image Dataset

This dataset contains the processed image subset used in the associated manuscript for concrete micro-surface classification and micro-crack detection. The images were acquired using a digital microscope under controlled laboratory conditions with three different water-to-cement ratios (0.35, 0.45, and 0.55), representing dense, intermediate, and porous surface structures. This uploaded version includes only the processed images (resized to 224×224 pixels) used directly for CNN training and evaluation. The dataset is organized into two main subsets: 1) Binary classification (micro-crack detection within dense surfaces): - cracked (150 images) - healthy (150 images) Total: 300 images 2) Multi-class surface classification: - dense (300 images) - intermedi (300 images) - porous (300 images) Total: 900 images Raw (uncropped) microscope images are not included in this release. The folder structure directly reflects class labels to ensure reproducibility and straightforward use in deep learning workflows. This version corresponds exactly to the data used in the submitted manuscript and is uploaded for peer review purposes. Citation request: If you use this dataset in your research, please cite the following associated preprint: Nogay, H. S. (2026). A Micro-Surface Concrete Image Dataset for Crack Detection and Texture Classification Using Deep Learning. Research Square. https://doi.org/10.21203/rs.3.rs-9594679/v1 Proper citation supports the continued development and maintenance of this dataset. This dataset is distributed under the Creative Commons Attribution 4.0 (CC BY 4.0) license. Users must provide appropriate credit when using this dataset.