Traffic Sign Recognition - Machine Learning Project

1. Problem Definition

In this project, a machine learning model will be developed to classify different traffic signs.
The automatic recognition of traffic signs provides a crucial application area for autonomous driving systems and road safety. The study aims to compare the performance of two different machine learning algorithms:

• A tree-based model (Random Forest or XGBoost).
• A deep learning-based model (Convolutional Neural Network, CNN).

---

2. Summary of the Dataset Used in the Study

The project will use the GTSRB - German Traffic Sign Recognition Benchmark dataset available on Kaggle (https://www.kaggle.com/datasets/meowmeowmeowmeowmeow/gtsrb-german-traffic-sign/data). This dataset contains the following details about various traffic signs:

• Total Number of Classes: 43 (e.g., stop sign, speed limit, warning signs, etc.).
• Total Number of Images: 50,000+.
• Image Dimensions: Each image will be normalized to 32x32 pixels.
• Dataset Structure: The dataset is organized into “train” and “test” folders, with each class having its own subfolder.

This dataset provides a sufficient amount of labeled data to train both machine learning and deep learning models.

Data Preprocessing:

• Images will be normalized to a size of 32x32 pixels.
• Data augmentation techniques (e.g., rotation, flipping) will be applied to improve model performance.

---

3. Machine Learning Algorithms and Evaluation Methods

Algorithms to be Used:

A - Tree-Based Model: Random Forest

B - Convolutional Neural Network (CNN)

Evaluation Methods:

• Accuracy: The accuracy rates of both models will be compared.

• Training Time: The training durations of the algorithms will be recorded.

• Confusion Matrix: An analysis will be conducted to determine which traffic signs are harder to recognize.

• Model Complexity: The complexity and computational cost of the tree-based model and CNN will be compared.

  4.DEMO