Traffic Sign Recognition with Neural Networks

This project aims to develop an AI system to identify traffic signs from photographs using neural networks, particularly with TensorFlow.

Introduction

Traffic sign recognition is a crucial component of autonomous driving systems, enabling vehicles to understand and interpret their surroundings from images captured by cameras. In this project, we utilize TensorFlow to build a neural network model that can classify different types of road signs based on input images.

Files

• traffic.py: Contains the implementation of functions to load data and build the neural network model.
• README.md: Documentation of experimentation process and observations.
• requirements.txt: List of required Python packages.

Usage

1. Download the dataset from the provided link and unzip it.
2. Move the dataset directory inside the traffic directory.
3. Install dependencies by running pip3 install -r requirements.txt.
4. Run the command python traffic.py <path_to_dataset> to train and evaluate the model.

Experimentation Process

For the experimentation process, I focused on tuning the architecture of the neural network model to achieve better accuracy in classifying traffic signs. Here's a summary of my experimentation:

• Convolutional Layers: I experimented with different numbers and sizes of convolutional layers to capture spatial features from input images effectively.
• Pooling Layers: Various pool sizes for pooling layers were tested to downsample the feature maps while retaining important information.
• Hidden Layers: I tried different numbers of hidden layers and neurons to find the optimal balance between model complexity and performance.
• Dropout: Dropout regularization was applied to prevent overfitting by randomly dropping neurons during training.
• Activation Functions: I explored different activation functions such as ReLU and tanh to introduce non-linearity into the model.
• Learning Rate: Adjustments to the learning rate were made to control the rate at which the model learns from the training data.

Observations

• Increasing the depth of the neural network by adding more layers led to better performance on the training set but also increased the risk of overfitting on the test set.
• Dropout regularization helped mitigate overfitting and improved the generalization ability of the model.
• The choice of activation function significantly influenced the convergence speed and final accuracy of the model.
• Experimenting with different hyperparameters requires careful tuning and experimentation to find the optimal configuration for the specific task.