metadata

title: Computer Mouse GAN Generator
emoji: 🖱️
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 3.50.0
app_file: app.py
pinned: false

Computer Mouse GAN Generator

This project uses a Deep Convolutional Generative Adversarial Network (DCGAN) to generate realistic images of computer mice. The model was trained on a dataset of computer mouse images with data augmentation to expand the training set.

Model Architecture

The model is based on the DCGAN architecture from the paper Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks.

Generator: 5 transpose convolutional layers with batch normalization
RGB color output (3 channels)
Trained with Weights & Biases monitoring
Training included data augmentation (flips, rotations, brightness/contrast adjustments)

Demo App

The Gradio interface allows you to:

Generate multiple computer mouse images at once
Set the number of images to generate (1-64)
Use a specific random seed for reproducible results
Toggle random seed generation for variety

Training Process

The model was trained on:

~300 original computer mouse images
Expanded to ~2,500 training samples through augmentation
Trained for 150+ epochs
Used CUDA acceleration on an RTX 3070

Examples

Generated images show a variety of computer mouse designs with different colors and shapes. Each image is completely new and generated from random noise - these mice don't exist in the real world!

Usage

Simply adjust the sliders and click "Generate Mice" to create new computer mouse designs.