G-Net: A TensorFlow-based Generative Adversarial Network

G-Net is an implementation of Generative Adversarial Networks (GAN) using TensorFlow. Designed to generate images of size 32x32 with 3 channels (RGB), G-Net provides real-time training statistics through TensorBoard integration.

Table of Contents

• Features
• Requirements
• Usage
• License

Features

• TensorFlow GAN: Built with TensorFlow, G-Net offers a robust platform for GAN model development.

• TensorBoard Integration: Get insights into your model's performance in real-time with TensorBoard logging.

• Customizable Noise Dimension: Tailor the noise dimension based on your specific requirements.

• Image Visualization: Track the evolution of the generated images through visualization after regular intervals.

Requirements

• TensorFlow (2.x recommended)
• numpy
• matplotlib
• pickle

Usage

1. Setup: Start by installing all required libraries:

   pip install tensorflow numpy matplotlib

2. Dataset Preparation: Use datasets in pickle format that can be reshaped into images of (32, 32, 3). The given function load_and_preprocess_data is set up for this purpose.

3. Training: To initiate model training, run:

   python <filename>.py

   (Replace <filename> with the name of the Python script containing the G-Net code).

4. Monitor with TensorBoard: Track the training progress visually using TensorBoard:

   tensorboard --logdir logs --reload_multifile true

5. Tweaking: Adjust the noise_dim to modify the noise dimension. For a different number of epochs, change the epochs variable.

License

G-Net is an open-source project under the MIT License.

---

Note: Ensure you include a license file if you reference it in the README.