roadsense :A Deep Learning-Based Framework for Automated Road Damage Detection & Pavement Condition Assessment

Gemini API
, , , ,
1

Build a professional AI-powered road damage detection web application called “RoadSense”.

Project purpose:
RoadSense detects and analyzes road damage using a custom-trained YOLO object detection model trained on the RDD2022 dataset.

The model detects four classes:

  1. D00 - Longitudinal Crack
  2. D10 - Transverse Crack
  3. D20 - Alligator Crack
  4. D40 - Pothole

Create a modern, professional computer-vision dashboard.

MAIN FEATURES:

  1. Landing Dashboard
  • Project name: RoadSense
  • Subtitle: AI-Powered Road Damage Detection & Pavement Condition Assessment
  • Modern dark/light professional design
  • Show statistics cards:
    Total Damages
    Potholes
    Cracks
    Road Health Score
  1. Image Analysis
  • Upload road image
  • Drag and drop interface
  • Analyze Image button
  • Display original image
  • Display detected image with bounding boxes
  • Show detected damage type
  • Show confidence score
  • Show number of detected damages

  1. Damage Analysis
    Create a table:
    Damage Type | Count | Average Confidence | Severity

  2. Severity Analysis
    Calculate:
    Low: 0-30
    Medium: 31-65
    High: 66-100

  3. Road Health Score
    Create a large circular 0-100 score.

Example:
61/100
Maintenance Recommended

  1. Road Condition
    Show:
    GOOD
    MODERATE
    POOR
    CRITICAL

  2. Recommendations
    Generate maintenance recommendations based on detected damage.

  3. Video Analysis
    Allow users to upload road videos.
    Process frames and detect road damage.

  4. Dashboard
    Include:

  • Damage distribution chart
  • Severity distribution
  • Detection confidence
  • Road health score
  • Total detected defects
  1. UI DESIGN
    Use a premium modern dashboard.
    Use:
  • cards
  • rounded corners
  • subtle shadows
  • responsive layout
  • clean typography
  • professional icons
  • charts
  • progress bars
  • modern navigation sidebar

Navigation:
Dashboard
Image Analysis
Video Analysis
Detection History
Analytics
About RoadSense

  1. Backend Integration

The application must be designed so that a custom YOLO model named:

best.pt

can be connected through an API endpoint.

Create a clear API service layer such as:

POST /predict

Input:
image

Output JSON:

{
“detections”: [
{
“class”: “D40”,
“name”: “Pothole”,
“confidence”: 0.91,
“bbox”: [x1,y1,x2,y2]
}
],
“severity”: “High”,
“road_health_score”: 61,
“recommendation”: “Maintenance Recommended”
}

Do not use a fake AI response for final inference.
Keep the model/API integration modular so the trained YOLO model can be connected later.

Make the UI production-quality and suitable for a college AI/Computer Vision project demonstration.