https://github.com/hayatiyrtgl/dcgan_mnsit

This code implements a DCGAN to generate realistic images of handwritten digits from the MNIST dataset, training the generator and discriminator alternatively for a set number of epochs.
https://github.com/hayatiyrtgl/dcgan_mnsit

gan gans generative-adversarial-network generator keras python tensorflow

Last synced: 3 days ago
JSON representation

This code implements a DCGAN to generate realistic images of handwritten digits from the MNIST dataset, training the generator and discriminator alternatively for a set number of epochs.

• Host: GitHub
• URL: https://github.com/hayatiyrtgl/dcgan_mnsit
• Owner: HayatiYrtgl
• License: mit
• Created: 2024-05-02T09:06:42.000Z (6 months ago)
• Default Branch: main
• Last Pushed: 2024-05-02T09:07:31.000Z (6 months ago)
• Last Synced: 2024-05-02T22:11:31.607Z (6 months ago)
• Topics: gan, gans, generative-adversarial-network, generator, keras, python, tensorflow
• Language: Python
• Homepage:
• Size: 33.2 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# DCGAN_mnsit

This code implements a DCGAN to generate realistic images of handwritten digits from the MNIST dataset, training the generator and discriminator alternatively for a set number of epochs.

This code implements a Deep Convolutional Generative Adversarial Network (DCGAN) to generate images of handwritten digits from the MNIST dataset. Here's a breakdown of the code:

1. **Imports**:

   - Necessary libraries are imported, including layers from Keras, TensorFlow, tqdm for progress bars, numpy, and functions to load and plot the MNIST dataset.

2. **Data Processing**:

   - The MNIST dataset is loaded and preprocessed.

   - Images are normalized and reshaped to be in the range [-1, 1] and with shape `(28, 28, 1)`.

3. **Discriminator Creation**:

   - A discriminator model is created using convolutional layers with LeakyReLU activations and dropout layers.

   - It outputs a single value representing the probability of the input being real.

   - The discriminator is compiled with binary cross-entropy loss and the Adam optimizer.

4. **Generator Creation**:

   - A generator model is created using transpose convolutional layers with BatchNormalization and ReLU activations.

   - It takes random noise as input and generates images with the same dimensions as MNIST digits.

   - The generator is compiled with binary cross-entropy loss and the Adam optimizer.

5. **Combining Generator and Discriminator to Form GAN**:

   - The generator and discriminator are combined sequentially to form the GAN model.

   - The discriminator's trainable parameter is set to `False` to prevent it from being trained during the generator training.

   - The GAN is compiled with binary cross-entropy loss and the Adam optimizer.

6. **Training**:

   - The GAN is trained for a specified number of epochs.

   - In each epoch, for each batch in the training dataset, the generator and discriminator are alternately trained.

   - The discriminator is first trained on real and fake images with corresponding labels.

   - Then, the generator is trained to generate images that fool the discriminator.

7. **Saving the Generator Model**:

   - After training, the generator model is saved to a file named "Generator.h5".

This code demonstrates the implementation of a DCGAN for generating realistic-looking handwritten digits.