Road Sign Detection using Convolutional Neural Networks (CNN)

Description

Road sign detection and classification play a crucial role in autonomous driving systems. This project utilises TensorFlow and Keras to create a CNN model that can classify road signs into various categories.

Dataset Directory Structure

Ensure the dataset is organised in the following directory structure:

ROAD-SIGN-DETECTION/
│
├── traffic Data/
│   ├── TEST/
│   ├── TRAIN/
│   └── VAL/
│
├── Classifying Road Signs.ipynb
├── labels.csv
├── model.h5
├── README.md
└── Splitting Dataset.ipynb

In this structure:

• The traffic Data directory contains subdirectories TEST, TRAIN, and VAL, which store the testing, training, and validation datasets respectively.
• Classifying Road Signs.ipynb is a Jupyter Notebook containing code for classifying road signs.
• labels.csv is a CSV file containing labels for the images in the dataset.
• model.h5 is the trained model saved in HDF5 format.
• README.md is the readme file containing project documentation and instructions.
• Splitting Dataset.ipynb is a Jupyter Notebook containing code for splitting the dataset into training and validation sets.

Splitting Dataset

To split the dataset, you can use the Splitting Dataset.ipynb notebook. This notebook uses the splitfolders library to split the dataset into training and validation sets with specified ratios.

Dependencies

Ensure you have the following dependencies installed:

• TensorFlow
• matplotlib
• pandas
• numpy
• PIL

You can install these dependencies via pip:

pip install tensorflow matplotlib pandas numpy Pillow 

Usage

1. Clone the repository:

git clone https://github.com/FunmiOlaitan/Road-Sign-Detection.git
cd Road-Sign-Detection

2. Load and Analyze Dataset:

   • The dataset is located in the traffic_Data directory.
   • Use the provided Python script to load and analyze the dataset, including class distribution and visualizations.

3. Preprocessing:

   • Images are preprocessed and augmented using the ImageDataGenerator from TensorFlow's Keras API.

4. Build Custom Model:

   • A custom CNN model is built using TensorFlow's Keras functional API. It consists of convolutional, max-pooling, flatten, dense, dropout, and softmax layers.

5. Train the Model:

   • The model is trained using the training data and validated using the validation data.
   • Training history (accuracy and loss) is visualised using Matplotlib.

6. Save the Model:

   • Once trained, the model is saved in the .h5 format for future use.

Customisation

You can customise the project by adjusting the following:

• Image dimensions and batch size for training.
• CNN model architecture and hyperparameters.
• Data augmentation techniques.
• Training epochs and learning rate.

Acknowledgements

Special thanks to the creators of the splitfolders library for simplifying dataset splitting and to the authors of the various libraries used in this project.

Support

For any issues or inquiries, feel free to open an issue in the repository. Your contributions and feedback are welcome to enhance the project further.

Credits:

For any questions or issues, please contact the author at Funmilolaolaitan@yahoo.com