Crack Detection

As part of the project, I built a Crack Segmentation system using YOLOv11 from Ultralytics , aiming to detect cracks in infrastructure images, Here's a snapshot of what I’ve done so far:

Dataset: Integrated via the Roboflow API, with over 3,800 images used for training after preprocessing and restructuring the data.

Model: Fine-tuned a YOLOv11 segmentation model (yolo11m-seg.pt) over 23 epochs for optimized performance.

Automation: Wrote Python scripts to handle dataset merging, labeling, and YAML config generation.

Results:

Box mAP50: 75.6%

Mask mAP50: 63%

This project has given me hands-on experience with:

Deep learning for semantic segmentation

Real-world computer vision applications

Model training with YOLOv11 on Google Colab

Using Roboflow for dataset management