In this project, we investigated various machine learning and deep learning models for traffic sign detection (TSD) to enhance road safety. Using the German Traffic Sign Recognition Benchmark (GTSRB) dataset, we aimed to identify the most effective models for accurately recognizing traffic signs.
Convolutional Neural Networks (CNNs), particularly the VGG16 architecture, demonstrated significant potential. VGG16 achieved the highest accuracy of 96.9%, which can be attributed to its ability to learn complex features directly from images.
In addition to VGG16, we evaluated several other models, obtaining the following test accuracies:
- Linear Discriminant Analysis (LDA): 71.33%
- Quadratic Discriminant Analysis (QDA): 64.06%
- Random Forest Classifier: 97.89%
- Support Vector Machine (SVM) Classifier: 60.83%
- Multilayer Perceptron (MLP) Classifier: 85.13%
- XGBoost Classifier: 83.42%
- Logistic Regression: 96.42%
- K-Nearest Neighbors (KNN): 90.76%
- CNN: 95.46%
- VGG16: 96.9%
Our analysis of precision, recall, and F1 scores highlighted the trade-offs between different models. VGG16 emerged as a highly promising choice for TSD due to its superior accuracy. However, further exploration of other models and techniques could yield additional improvements.
This project underscores the potential of deep learning in enhancing traffic sign detection systems, contributing significantly to improved road safety.