Synthetic-Image-Generation

This repository deals with synthetic image generation using Deep Convoluted General Adverserial Network (DCGAN). The main dataset used in this repository is the pre-defined Fashion MNIST dataset.

Installations

This project uses Python 3.x, so a suitable python version (>=3.6) is advisable.

The project uses the following frameworks:-

• tensorflow
• numpy
• plot_utils
• matplotlib
• tqdm
• IPython

Local Machine installation

If you do not have a working installation of these packages, the easiest way to install is using pip

pip install <package name >

or conda:

conda install <package name>

Project Motivation

The Max-min strategy of GAN was very interesting which led me to this project. There are two neural networks competing with each other which results in a better result.

Summary of the Result

This is a Computer Vision Project which uses deep convoluted neural network to generate Images that are close to the real dataset.

The Discriminator and Generator neural networks are trained and the quality of the result after each epochs are shown:-

1/10 epoch:-

2/10 epoch:-

3/10 epoch:-

4/10 epoch:-

5/10 epoch:-

6/10 epoch:-

7/10 epoch:-

8/10 epoch:-

9/10 epoch:-

10/10 epoch:-

File Descriptions

There are two files plot_utils.py and DCGAN.inpyb

1. The show function of plot_utils.py is used to visulize the images generated by the Generator after each epochs.

2. The file DCGAN.inpyb contains the code for the Synthetic image generation using Deep Convoluted Generative Adverserial Neural Network

Source code

You can check the latest sources with the command:

https://github.com/deadshotsb/Synthetic-Image-Generation.git

How to interact

Download/clone the repository and run the file provided in the repository.

In case of harware/Package failure you can try out Google Colab .

License

The project is under Apache 2.0 License