snehadm/CNN

🧠 CIFAR-10 Visual Intelligence

A state-of-the-art Flask-based web application powered by a custom Convolutional Neural Network (CNN). This application provides real-time image classification for the 10 object classes in the CIFAR-10 dataset.

---

🚀 Features

• High-Accuracy Classification: Optimized CNN architecture for small image recognition.
• Real-time Inference: Quick results with confidence percentage bars.
• Containerized Deployment: Ready-to-use Docker environment.
• Premium UI/UX: Modern dark-mode interface with glassmorphism aesthetics.
• API Endpoint: Accessible /predict POST endpoint for programmatic use.

---

🏗️ Model Architecture

The underlying model is a Lightweight CNN optimized for CPU performance while maintaining high accuracy.

Layer Type	Configuration	Filters/Units
Input	Image	32 x 32 x 3
Augmentation	Flip, Rotation, Zoom	-
Conv Block 1	(Conv2D x 2) + MaxPool + Dropout	32 filters
Conv Block 2	(Conv2D x 2) + MaxPool + Dropout	64 filters
Conv Block 3	Conv2D + MaxPool + Dropout	128 filters
Dense Head	Flatten + Dense + BN + Dropout	128 units
Output	Dense (Softmax)	10 classes

---

📊 Dataset: CIFAR-10

The CIFAR-10 dataset consists of 60,000 32x32 color images in 10 classes, with 6,000 images per class.

Classes: airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck

---

🛠️ Installation and Usage

Local Installation

1. Clone the repository:

   git clone https://github.com/your-username/cifar10-app.git
   cd cifar10-app
   

2. Set up environment:

   python -m venv venv
   # Windows:
   .\venv\Scripts\activate
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Run the application:

   python app.py
   

Docker Deployment

docker build -t cifar10-app .
docker run -p 5000:5000 cifar10-app

---

📦 Project Structure

├── app.py              # Flask Web Server
├── train.py            # Model Training Logic
├── predict.py          # Command Line Inference
├── Dockerfile          # Container Configuration
├── .gitignore          # Git Exclusion Rules
├── requirements.txt    # Python Dependencies
├── static/             # CSS, Images, and Uploads
├── templates/          # HTML Frontend
└── samples/            # Test Image Collection

---

⚠️ Troubleshooting

• Dependency Conflict: If numpy or scipy causes import loops, ensure you have the pinned versions: numpy<2.0 and scipy<1.14.
• OneDNN Warnings: These are normal for Intel/AMD CPUs and do not affect functionality.

---

Created with ❤️ for Visual Intelligence.