Convolution Feature Map Visualizer 🏞️

A Streamlit-based web app that allows you to visualize the feature maps generated by a convolution layer (Conv2d) in PyTorch. This tool is perfect for understanding how convolutional filters interact with images and exploring deep learning fundamentals.

Features

• Upload an image in PNG, JPG, or JPEG format.

• Adjust convolution parameters using an interactive sidebar:

  • Out Channels: Number of convolutional filters.
  • Kernel Size: Size of each convolutional filter.
  • Stride: Step size for sliding the kernel over the image.
  • Padding: Zero-padding added to the input image.

• Visualize each feature map in a grid layout.

• Developed entirely with PyTorch, Torchvision, PIL, Matplotlib, and Streamlit.

Demo

Installation

1. Clone the repository:

git clone https://github.com/RohanSardar/Conv2dVisualization
cd Conv2dVisualization

2. Create and activate a virtual environment (optional but recommended):

python -m venv venv
source venv/bin/activate  # Linux/Mac
venv\Scripts\activate     # Windows

3. Install dependencies:

pip install -r requirements.txt

Usage

1. Run the Streamlit app:

streamlit run app.py

2. Upload an image through the sidebar.
3. Adjust convolution parameters.
4. Click "Process Image" to visualize the feature maps.

Convolution Parameters

• Out Channels: Controls how many different features the convolution layer extracts.
• Kernel Size: Determines the size of the filter used to scan the image.
• Stride: Determines how much the filter moves at each step.
• Padding: Adds zeros around the image to control output size.

Example Output

Each filter produces a feature map showing activated patterns from the input image. This visualization helps understand how convolution layers detect edges, textures, and more complex patterns.

Built With

• Python
• Streamlit
• PyTorch
• Torchvision
• PIL / Pillow
• Matplotlib

Author

Developed by Rohan Sardar Project completed on 2nd December 2025.