A deep learning project that uses Convolutional Neural Networks (CNNs) to detect road obstacles (potholes, speed bumps, etc.) from camera images — useful for autonomous driving systems and smart transportation.

📦 Dataset

The dataset is automatically downloaded from Kaggle: Road Obstacles Detection Dataset

It contains labeled images of:

• Normal roads
• Potholes
• Speed bumps

🧠 Model

This project uses TensorFlow/Keras to build and train a CNN with the following goals:

• Classify road images into obstacle types
• Achieve lightweight inference suitable for real-time use
• Handle imbalanced datasets with class_weight

Architecture Highlights:

• Input shape: 224×224 images
• Standard CNN layers with ReLU and max-pooling
• Dense output layer for classification

🛠️ Requirements

Make sure you have the following packages installed:

pip install tensorflow matplotlib scikit-learn kagglehub