Skip to content

README.md

Latest commit

 

History

History
103 lines (70 loc) · 2 KB

README.md

File metadata and controls

103 lines (70 loc) · 2 KB
Raw

🧥 Fashion MNIST Classifier

This project is a PyTorch-based image classifier for the Fashion MNIST dataset using a RegNetY architecture from the timm library. It leverages Albumentations for image preprocessing and visualizes training and validation loss.

📦 Dataset used: Fashion MNIST PNG Dataset on Kaggle
🧠 Model: regnety_006 from timm

📊 Sample Loss Plot

After training, the script will save the following plot to assets/loss_plot.png:

Training and Validation Loss

🚀 Installation

  1. Clone the repository

    git clone https://github.com/anto18671/vision-classifier.git
cd vision-classifier

  2. Create a virtual environment (optional but recommended)

    python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows

  3. Install dependencies

    pip install -r requirements.txt

📁 Dataset Structure

Download and unzip this dataset into the data/ folder, so that it looks like:

vision-classifier/
│
├── data/
│   ├── train/
│   │   ├── 0/
│   │   ├── 1/
│   │   └── ...
│   └── val/
│       ├── 0/
│       ├── 1/
│       └── ...

🏁 Run the Training

Simply run:

python train.py

This will:

  • Train the model for 24 epochs
  • Save the training/validation loss plot to assets/loss_plot.png

📦 Requirements

Minimal requirements.txt:

torch
torchvision
timm
albumentations
opencv-python
matplotlib
tqdm

You can create this file with:

pip freeze > requirements.txt

📄 License

MIT License — see LICENSE