Fresh Concrete Surface Crack Detection Dataset: 659 Field Images with Pixel and Instance Annotations

Published: 15 September 2026| Version 2 | DOI: 10.17632/r8zrzk5w8j.2
Contributors:
,

Description

Version 2 Update: - Reorganized dataset structure with explicit train/valid/test splits - Added separate folders for YOLOv8 polygon annotations (labels_yolo/) and U-Net binary masks (masks_unet/) - Included data.yaml configuration file for direct use with Ultralytics YOLOv8 - Added comprehensive README.txt with dataset documentation - Total: 659 images (461 train / 132 validation / 66 test) with both pixel-level and instance-level annotations

Files

Institutions

Categories

Computer Vision, Civil Engineering, Image Segmentation, Quality Control, Construction Management, Concrete Technology, Surface Crack, Deep Learning

Licence