🖤 MNIST Digit Classifier – PyTorch

A Convolutional Neural Network (CNN) built using PyTorch to classify handwritten digits from the MNIST dataset.
This project loads a trained model and predicts digits from custom images.

---

📌 Features

• Dataset: MNIST (28×28 grayscale digit images)
• Model Architecture:
  • 3 Convolutional Layers + ReLU activation
  • Flatten layer
  • Fully Connected Layer (10 outputs for digits 0–9)
• Loss Function: CrossEntropyLoss
• Optimizer: Adam
• Device Support: GPU (CUDA) or CPU

---

📂 Project Structure ├── data/ # MNIST dataset (auto-downloaded) ├── img_1.jpg # Sample image for prediction ├── model_state.pt # Saved model weights ├── torchnn.py # Main script └── README.md # Documentation

---

🚀 Installation

1️⃣ Clone the repository : --bash : git clone https://github.com/RIVALHIDE/pytorch-minst.git cd RIVALHIDE

2️⃣ Create a virtual environment : --bash: python -m venv .venv

3️⃣ Activate the environment :

Windows: --bash : .venv\Scripts\activate

4️⃣ Install dependencies : --bash : pip install torch torchvision pillow

▶️ Usage :

Run the prediction script: --bash : python torchnn.py

Example Output: Using device: cpu Total parameters: 365514 Predicted Digit: 7