Road Anomaly, Pothole & Road Defect Detection Ecosystem (YOLOv8)

State-of-the-art fine-tuned YOLOv8 deep neural models for real-time automated detection of road surface hazards, potholes, severe structural cracks, and pavement anomalies across urban and highway corridors.

---

📦 Model Suite & Checkpoints

---

🎯 Benchmark Performance

---

🚀 Quick Start with Ultralytics

1. Dedicated Pothole & Road Anomaly Inference (PyTorch)

from ultralytics import YOLO

# Load 7-class primary model
model = YOLO("pothole_yolov8.pt")

# Or load the PotBot dedicated high-capacity pothole model
# model = YOLO("potbot_yolov8m.pt")

# Run real-time inference on a dashcam video or camera stream
results = model.predict(source="road_video.mp4", conf=0.30, save=True)

for r in results:
    for box in r.boxes:
        cls_id = int(box.cls[0])
        cls_name = model.names[cls_id]
        conf = float(box.conf[0])
        print(f"Detected {cls_name} with confidence {conf:.2f}")

2. High-Performance ONNX Runtime (CPU / Edge Deployment)

import cv2
import numpy as np
import onnxruntime as ort

session = ort.InferenceSession("pothole_yolov8.onnx", providers=['CPUExecutionProvider'])
input_name = session.get_inputs()[0].name

# Preprocess image to 640x640 RGB float32
img = cv2.imread("road_scene.jpg")
blob = cv2.dnn.blobFromImage(img, 1/255.0, (640, 640), swapRB=True)

outputs = session.run(None, {input_name: blob})
print("ONNX Inference Output Shape:", outputs[0].shape)

---

🏷️ Citations & Acknowledgments

• RAD & Indian Roads Dataset: Municipal pavement patrol data (Chennai Corporation).
• CRDDC2022: Global Road Damage Detection Challenge 2022 benchmark dataset.
• PotBot System: AI-Powered Pothole Detection & Stereo-Vision Prototype (RidaArshad / Rohan-Aroli).
• Ultralytics: YOLOv8 framework and training algorithms.