https://github.com/shrikantnaidu/face-generation

Face Generation with DCGAN
https://github.com/shrikantnaidu/face-generation

deep-learning face-generation generative-adversarial-network

Last synced: 26 days ago
JSON representation

Face Generation with DCGAN

• Host: GitHub
• URL: https://github.com/shrikantnaidu/face-generation
• Owner: shrikantnaidu
• License: mit
• Created: 2024-04-27T10:58:20.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2024-04-28T07:53:10.000Z (9 months ago)
• Last Synced: 2024-11-11T21:31:14.314Z (3 months ago)
• Topics: deep-learning, face-generation, generative-adversarial-network
• Language: Jupyter Notebook
• Homepage:
• Size: 783 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Face-Generation

This project trains a Deep Convolutional Generative Adversarial Network (DCGAN) on a dataset of faces to generate new, realistic images of human faces.

## Overview

- dlnd_face_generation.ipynb - Jupyter notebook containing the project code and explanations

- problem_unittests.py - Unit tests for model solutions

- processed_celeba_small/ - Directory of preprocessed CelebA face images

## Dependencies

- Python 3.x

- PyTorch

- Torchvision

- Matplotlib

- Numpy

## Usage

To run the Jupyter notebook:

1. Clone this repository

2. Install dependencies

3. Run jupyter notebook in the repository directory

4. Open dlnd_face_generation.ipynb

The notebook contains sections on:

- Loading and visualizing the CelebA dataset

- Defining discriminator and generator models

- reating loss functions and optimizers

- Training the GAN

- Visualizing generated samples

- Preprocessed CelebA images are provided in processed_celeba_small/. New images can be generated after training the GAN.

## Acknowledgements

The CelebA dataset is from http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html

This project was completed as part of the Udacity Deep Learning Nanodegree. The starting code and dataset were provided by Udacity.

## License

The contents of this repository are covered under the MIT License.