Digit Recognizer & Image Classification (CNN)

This repository contains two folders of Convolutional Neural Networks (CNNs) applied to different classification tasks.

📁 Project Structure

Digit-Recognizer/
│
├── ai_vs_real/
│   └── source.py
│
├── digit_recogn/
│   └── source.py

---

🔹 1. Pixel CNN (Digit Recognizer)

• CNN model applied to pixel-based data (numerical vectors)
• Does not use .jpg images, but preprocessed pixel values
• Similar to classic digit recognition problems (e.g. MNIST)

Features:

• Input: pixel vectors
• Model: simple CNN
• Task: digit classification

---

🔹 2. Image CNN (Real vs AI Art)

• CNN model applied to real images (.jpg)

• Classifies images into:

  • 0 → real art
  • 1 → AI-generated art

Features:

• Input: RGB images
• Preprocessing: resizing, augmentations
• Model: multi-layer CNN
• Metric: F1-score

---

⚙️ Technologies Used

• Python
• PyTorch
• Pandas
• NumPy
• PIL

---

🚀 How to Run

1. Install dependencies:

pip install torch torchvision pandas pillow scikit-learn

2. Run the desired script:

python source.py

---

🧠 Purpose

This repository explores:

• the difference between numerical and image-based inputs
• how CNNs behave in different contexts
• a full pipeline: dataset → model → evaluation

---

📌 Note

This code is intended for learning and experimentation purposes and is not fully optimized for production use.

---