Car Brand Classifier.
An image classification machine learning pipeline designed to identify car brands with high accuracy. The project leverages transfer learning to recognize intricate automotive features.
Key Features
The pipeline is designed to be robust and portable, handling everything from raw image input to confident predictions.
Preprocessing Engine
Automatically preprocesses input images (resizing, scaling) for optimal model inference.
Confidence Visualization
Provides classification confidence percentage visualization via Matplotlib.
Docker Containerization
Fully containerized via Docker for portable and consistent inference across environments.
Model Architecture
The core of the classifier relies on EfficientNetB0, a highly optimized convolutional neural network. By fine-tuning this pre-trained architecture, the model learns brand-specific features without requiring massive amounts of computational resources.
- Architecture: EfficientNetB0 (Fine-tuned)
- Dataset: 11,000+ images across 50 distinct car brands
- Framework: Built and trained with TensorFlow & Keras
- Computer Vision: OpenCV utilized for advanced image processing operations