📸 Embedded Camera Image Classification System

An end-to-end production-grade image classification system designed for edge-device scenarios. This project demonstrates the integration of Deep Learning, Asynchronous API design, and Clean Architecture to create a scalable vision service.

---

🎯 Project Objective

The goal of this system is to provide a lightweight inference API that can be deployed on low-resource devices (like Raspberry Pi or Jetson Nano) to classify images captured by embedded cameras. It balances the trade-off between model accuracy and inference latency.

🚀 Key Features

• Lightweight CNN: A custom-built Convolutional Neural Network trained on the CIFAR-10 dataset.
• Asynchronous API: Built with FastAPI to handle non-blocking I/O, allowing the system to process multiple image streams simultaneously.
• OOP Design: Implemented using a Service-Oriented Architecture (SOA) to ensure separation of concerns between image processing, inference, and transport layers.
• Robust Validation: Strict input validation for file types (JPG, PNG) and size limits (10MB) to prevent system crashes.
• Concurrency Benchmarking: A dedicated experiment suite to compare Sequential, Threaded, and AsyncIO performance.
• Automated Pipeline: Scripts provided for training, evaluation (Confusion Matrix), and live demo.

---

🛠️ Tech Stack

• Language: Python 3.10+
• Deep Learning: PyTorch, Torchvision
• API Framework: FastAPI, Uvicorn
• Image Processing: Pillow (PIL), NumPy
• Evaluation: Matplotlib, Seaborn, Scikit-learn
• Testing: HTTPX, Pandas

---

📁 Project Structure

project/
├── app/
│   ├── main.py                 # API Entry point
│   ├── routes/
│   │   └── predict.py          # Prediction endpoints
│   ├── services/
│   │   ├── inference_service.py # Model loading and prediction logic
│   │   └── image_processor.py   # Preprocessing (Resize/Normalize)
│   └── utils/
│       ├── decorators.py        # @log_execution and @timing decorators
│       └── validation.py        # Image file and size validators
├── ml/
│   ├── model.py                # CNN Architecture definition
│   ├── train.py                # Training script
│   ├── evaluate.py             # Accuracy and Confusion Matrix script
│   └── model.pt                # Saved trained weights
├── experiments/
│   └── concurrency_test.py     # Performance benchmarking tool
├── demo/
│   └── demo.py                 # Client-side demonstration script
├── requirements.txt            # Project dependencies
└── Dockerfile                  # Production containerization