📚 About & Help

Learn how to use the Road Hazard Detection System

About Road Hazard Detection

The Road Hazard Detection System is an advanced AI-powered platform that automatically detects road hazards (bumps and potholes) from sensor data. Using state-of-the-art machine learning models, it analyzes accelerometer and gyroscope readings to identify road surface anomalies with high accuracy.

Key Features

AI-Powered

Advanced TCN-KAN-LSTM neural network model for accurate predictions

GPS Mapping

Automatically maps hazards to exact GPS coordinates

Analytics

Comprehensive statistics and insights about road conditions

Fast Processing

Real-time analysis of large CSV datasets

Secure

Your data is processed securely and not permanently stored

Web-Based

Access from any device with a web browser

Model Performance

97.68%
Accuracy
3
Classes
108
Features

Supported Predictions

  • Normal: Road surface is in good condition
  • Bump: Speed bump or raised road surface detected
  • Pothole: Pothole or depression in road surface detected

How It Works

1. Data Collection

The system accepts CSV files containing sensor data from vehicles. This typically includes:

  • Accelerometer readings (X, Y, Z axes)
  • Gyroscope readings (X, Y, Z axes)
  • GPS coordinates (latitude, longitude)
  • Timestamps

2. Data Preprocessing

The uploaded data is cleaned and prepared:

  • Removes non-numeric columns
  • Extracts GPS coordinates
  • Handles missing values
  • Normalizes sensor readings

3. Feature Extraction

The system extracts 108 robust features from the raw sensor data:

  • Signal derivatives and rolling statistics
  • Energy and magnitude features
  • Frequency domain features
  • Statistical moments

4. Model Inference

The trained TCN-KAN-LSTM model processes the features and generates predictions:

  • Temporal Convolutional Network (TCN) for temporal patterns
  • Kolmogorov-Arnold Network (KAN) for non-linear transformations
  • LSTM for sequence dependencies

5. Result Processing

Results are mapped to GPS coordinates and presented with confidence scores:

  • Individual predictions with coordinates
  • Confidence scores for each prediction
  • Summary statistics
  • Interactive map visualization

Video Tutorial

Watch this step-by-step tutorial to learn how to use the Road Hazard Detection System:

Complete Tutorial - How to Use the System

Duration: Check video | Level: Beginner

This video covers everything you need to know to get started with the system, from preparing your data to analyzing results.

Video Topics Covered:

  • ✅ Preparing your CSV data
  • ✅ Uploading files to the system
  • ✅ Running predictions
  • ✅ Understanding the results
  • ✅ Interpreting the map visualization
  • ✅ Exporting results
📝 Note: The video file (how_to.mp4) should be placed in the root directory of the project.

Getting Started

Step 1: Prepare Your Data

Create a CSV file with your sensor data. Your CSV should include:

  • Latitude column (named: latitude, lat, or Latitude)
  • Longitude column (named: longitude, lng, lon, or Longitude)
  • Sensor data columns (accelerometer, gyroscope, etc.)

Example CSV structure:

latitude,longitude,accelerometer_x,accelerometer_y,accelerometer_z,gyroscope_x,gyroscope_y,gyroscope_z 40.7128,-74.0060,0.1,0.2,9.8,0.01,0.02,0.03 40.7129,-74.0061,0.15,0.25,9.85,0.015,0.025,0.035 40.7130,-74.0062,0.12,0.22,9.82,0.012,0.022,0.032

Step 2: Upload Your File

  1. Click on the upload area or drag and drop your CSV file
  2. Select your CSV file from your computer
  3. Wait for the file to be selected

Step 3: Run Prediction

  1. Click the "Predict Road Conditions" button
  2. Wait for the analysis to complete (usually takes a few seconds)
  3. View results on the map and in the statistics panel

Step 4: Analyze Results

  • View hazards marked on the interactive map
  • Check confidence scores for each prediction
  • Review summary statistics
  • Identify high-risk areas
💡 Tip: For best results, ensure your CSV has at least 50 data points and includes accurate GPS coordinates.

Frequently Asked Questions

What file formats are supported?
Currently, only CSV (Comma-Separated Values) files are supported. Ensure your file is properly formatted with headers in the first row.
What is the maximum file size?
There is no strict file size limit, but very large files (>100MB) may take longer to process. For optimal performance, keep files under 50MB.
How accurate is the model?
The model achieves 97.68% accuracy on the validation dataset. However, accuracy may vary depending on data quality, sensor calibration, and road conditions. Always verify critical predictions independently.
Is my data stored on the server?
No. Your CSV files are processed in memory and not permanently stored. Results are returned to you immediately and not retained unless you explicitly save them. Server logs are kept for 30 days for troubleshooting.
What sensor data do I need?
At minimum, you need accelerometer data (X, Y, Z axes). Gyroscope data is optional but improves accuracy. GPS coordinates (latitude/longitude) are required for mapping hazards to locations.
Can I use this for commercial purposes?
The model and service are proprietary. Commercial use requires a license agreement. Please contact us for licensing information.
What does "confidence score" mean?
The confidence score (0-1) indicates how certain the model is about its prediction. A score of 0.95 means the model is 95% confident in the prediction. Higher scores are generally more reliable.
How do I report issues or provide feedback?
Please contact us at support@connected-roads.local with any issues, feedback, or feature requests. We appreciate your input to help improve the service.
← Back to Home