Traffic-Signs-Recognition-using-CNN-Keras

A web application built with Flask and TensorFlow/Keras that classifies German traffic signs from user-uploaded images. The model is trained on the German Traffic Sign Recognition Benchmark (GTSRB) dataset.

Features

• Web Interface: Simple frontend (likely defined in templates/index.html) for uploading traffic sign images [1].
• Image Classification: Predicts the traffic sign category from the uploaded image using a trained Convolutional Neural Network (CNN) [1].
• 43 Classes: Recognizes 43 different types of German traffic signs [1][3].
• Confidence Score: Provides the confidence level of the prediction [1].
• API Endpoint: Includes a /predict endpoint for programmatic access [1].

Dataset

This project utilizes the German Traffic Sign Recognition Benchmark (GTSRB) dataset [3].

• Contains images of 43 different traffic sign classes [3].
• Images are preprocessed to 30x30 pixels with 3 color channels (RGB) for model input [3][1].

Model Architecture

The classification model (traffic_sign_model.h5) is a Convolutional Neural Network (CNN) built using TensorFlow/Keras. The architecture, based on the training notebook (gtsrb-cnn.ipynb), includes [3]:

• Multiple Conv2D layers with ReLU activations for feature extraction.
• MaxPool2D layers for down-sampling.
• BatchNormalization layers for stabilizing training.
• A Flatten layer to prepare features for dense layers.
• Dense layers with ReLU activation.
• Dropout for regularization to prevent overfitting.
• A final Dense layer with 43 units and softmax activation for multi-class classification [3].

(Note: A PyTorch model definition (model.py) is also present, but the deployed application (app.py) uses the TensorFlow/Keras model saved in traffic_sign_model.h5) [1][2][3].

Technology Stack

• Backend: Python, Flask [1]
• Machine Learning: TensorFlow/Keras [1][3]
• Libraries: NumPy, Pillow (PIL), Flask-CORS [1][3]
• (Training/Analysis): Pandas, Matplotlib, Scikit-learn, Seaborn, Jupyter Notebook [3]

Installation

1. Clone the repository:

   git clone <(https://github.com/ansarimzp/Traffic-Signs-Recognition-using-CNN-Keras)>
   cd <REPOSITORY_DIRECTORY>
   

2. Create and activate a virtual environment (recommended):

   python -m venv venv
   # On Windows:
   # venv\Scripts\activate
   # On macOS/Linux:
   # source venv/bin/activate
   

3. Install dependencies:
   • Ensure you have Python 3.x installed.
   • Create a requirements.txt file with the following content:

     Flask
     Flask-Cors
     numpy
     tensorflow
     Pillow
     # Add other dependencies if needed (e.g., gunicorn for deployment)
     

   • Install the requirements:

     pip install -r requirements.txt
     

4. Ensure Model File: Make sure the traffic_sign_model.h5 file is in the same directory as app.py, or update the path in app.py accordingly [1].

Usage

1. Run the Flask application:

   python app.py
   

2. Access the web interface: Open your web browser and navigate to http://127.0.0.1:5000/ (or the address provided in the console output) [1].
3. Upload an image: Use the file upload form to select a traffic sign image and submit it.
4. View prediction: The application will display the predicted traffic sign label and the confidence score [1].

API Usage:

You can also send a POST request directly to the /predict endpoint with the image file: