Road Sign Classification
This GitHub project is dedicated to road sign classification, a pivotal field in computer vision. By leveraging high-performing deep learning models and advanced methodologies, this project aims to identify and categorize road signs with precision. Using the GTSRB dataset and robust techniques, it provides innovative solutions for driver assistance systems (ADAS) and autonomous driving.
DEEP LEARNINGCONVOLUTIONAL NEURAL NETWORK
1 min read
Main Features
Data Augmentation: Enriched dataset variability through rotation, flipping, zooming, and cropping to improve model generalization.
Regularization Techniques: Implementation of Dropout to reduce overfitting and Batch Normalization to enhance training efficiency and stability.
Optimized Model Development: Deployment of both a standard CNN and a ResNet-inspired model with fine-tuned hyperparameters for maximum accuracy.
Adversarial Testing: Evaluation under adverse conditions such as noise, motion, and occlusion to ensure model robustness.
Technology Stack
Frameworks: TensorFlow, Keras
Dataset: German Traffic Sign Recognition Benchmark (GTSRB)
Techniques: CNN, ResNet-inspired architectures, Grad-CAM for visual interpretation
Data Preprocessing: Resizing, normalization, augmentation, and testing under adverse conditions