This project aims to develop an automated pothole detection system using deep learning-based object detection models (YOLO).
The goal is to detect and locate potholes on road surfaces from real-world images, which can help in road maintenance and safety management.
The dataset was collected, labeled, and augmented using Roboflow, and the model was trained using YOLOv11 Object Detection.
- Source: Custom dataset labeled using Roboflow
- Classes:
potholeobjects(other road elements)
| Dataset Type | Number of Images |
|---|---|
| Training Set | 1083 |
| Validation Set | 146 |
| Test Set | 66 |
| Total | 1295 |
- Model Type: YOLOv11 (Accurate)
- Training Platform: Roboflow Train
- Checkpoint: Pretrained on MS COCO (Best 47.0% mAP)
- Epochs Trained: 100
- Image Resolution: 640x640
- Augmentations Applied:
- Blur
- Rotation
- Brightness & Contrast Variation
- Noise Addition
- Horizontal & Vertical Flip