README.md

metadata

title: CIFAR-10 Image Classifier
emoji: 🚀
colorFrom: blue
colorTo: red
sdk: gradio
sdk_version: 4.44.1
app_file: app.py
pinned: false

CIFAR-10 Image Classifier

This is a Gradio interface for a convolutional neural network trained on the CIFAR-10 dataset. The model can classify images into 10 different object categories: Airplane, Automobile, Bird, Cat, Deer, Dog, Frog, Horse, Ship, and Truck.

Model Architecture

The neural network has the following architecture (based on the PyTorch CIFAR-10 Tutorial):

• Two convolutional layers with ReLU activation and max pooling
• Three fully connected layers
• Designed for 32x32 RGB input images

Input → Conv2d(3, 6, 5) → ReLU → MaxPool2d(2, 2) → 
Conv2d(6, 16, 5) → ReLU → MaxPool2d(2, 2) → 
Flatten → Linear(16*5*5, 120) → ReLU → 
Linear(120, 84) → ReLU → Linear(84, 10) → Output

Features

• Interactive image classification interface with modern UI
• Example images for each CIFAR-10 class
• Real-time predictions with probability scores
• Support for custom image uploads
• Built-in drawing tool for creating test images
• Responsive design with gradient backgrounds and animations
• Automatic image preprocessing (resize to 32×32)

How to Use with Your Existing Model

1. Place your trained PyTorch model file in the app directory and name it model.pth

2. Ensure your model uses the same architecture as defined in the Net class

3. Install the required dependencies:

   pip install -r requirements.txt
   

4. Run the application:

   python app.py
   

5. Access the interface at http://localhost:7860 (or the URL provided in the terminal)

Image Input Capabilities

The model can handle various types of image inputs:

Supported Image Formats

• JPG, PNG, BMP, TIFF, and other common image formats
• Color images (RGB with 3 channels)
• Any resolution (automatically resized to 32×32 pixels)

Robustness Features

• Resolution Independence: Works with images of any size (resized to 32×32)
• Color Preservation: Maintains RGB color information
• Contrast Handling: Works with both high and low contrast images
• Noise Tolerance: Can handle some image noise
• Rotation Tolerance: Some tolerance to slight rotations

Best Practices for Good Results

1. Center the object in the image area
2. Use clear contrast between the object and background
3. Fill most of the image area with the object
4. Avoid excessive noise or artifacts
5. Ensure the object is clearly visible

CIFAR-10 Classes

The model classifies images into these 10 categories:

1. Airplane - Aircraft flying in the sky
2. Automobile - Cars and vehicles on the road
3. Bird - Flying or perched birds
4. Cat - Domestic cats and felines
5. Deer - Wild deer and similar animals
6. Dog - Domestic dogs and canines
7. Frog - Amphibians like frogs
8. Horse - Horses and similar animals
9. Ship - Boats and ships on water
10. Truck - Trucks and heavy vehicles

Deployment to Hugging Face Spaces

This application can be deployed to Hugging Face Spaces by:

1. Creating a new Space on Hugging Face
2. Uploading these files to the repository
3. Setting the SDK to "Gradio"
4. Adding the requirements in the requirements.txt file
5. Uploading your model.pth file

The Space will automatically run the app.py file as the entry point.

Example Usage

The interface comes with simple example images representing each CIFAR-10 class. You can:

1. Click on any example image to load it
2. Upload your own image using the file browser
3. Draw an image using the built-in sketch tool
4. View the classification probabilities for each class

Try these examples:

• Simple drawings of objects from each class
• Photos of objects that match the CIFAR-10 categories
• Images with varying styles and backgrounds

Technical Details

This implementation is based on the PyTorch CIFAR-10 tutorial and includes:

• Gradio interface with custom CSS styling
• Image preprocessing pipeline (resize to 32x32)
• Softmax probability output
• Example generation for demonstration
• Model loading functionality for your trained weights

The prediction function:

model.eval()
with torch.no_grad():
    output = model(input_tensor)
    probabilities = F.softmax(output, dim=1)

The UI features:

• Animated gradient background
• Glass-morphism design elements
• Responsive layout for all screen sizes
• Interactive buttons with hover effects
• Clean, modern typography

Requirements

• Python 3.6+
• PyTorch >= 1.7.0
• TorchVision >= 0.8.0
• Gradio >= 4.0.0
• Pillow >= 8.0.0
• NumPy >= 1.19.0

Install with:

pip install -r requirements.txt

Troubleshooting

If you encounter issues:

1. Ensure your model.pth file is in the same directory as app.py
2. Verify that your model uses the same architecture as defined in the Net class
3. Check that all required dependencies are installed
4. Make sure you're using a compatible version of Python (3.6+)

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference