This project implements the Boundary Equilibrium Generative Adversarial Network (BEGAN) architecture. BEGAN is a type of Generative Adversarial Network (GAN) that aims to balance the generator and discriminator during training. Learn more here.
-
Flexible Image Dimensions: The architecture can dynamically adjust and train based on the input image size, enabling versatile usage across different datasets without the need for excessive pre-processing.
-
Upscaling Module: Notably, this project comes equipped with a robust upscaling module. You have the flexibility to train your model in low resolution, optimizing the training time. Once done, employ the upscaling utility to enhance both image and video resolutions. This unique feature not only streamlines the process but also optimizes resource utilization and efficiency.
-
Robust Codebase: Built with stability and efficiency in mind, the code is resilient against common pitfalls and challenges in GAN training, ensuring consistent and reliable performance.
-
Comprehensive Tool: Unlike many other GAN codes, this isn't just for training a GAN. This project covers the entire GAN utilization spectrum – from training to image generation and even video generation.
-
Intuitive Use: Whether you're a GAN veteran or just starting out, the project is designed to be user-friendly. Command-line arguments make operations straightforward, allowing easy toggling between training, image generation, and video creation.
-
Detailed Documentation: Along with this README, a well-structured Wiki provides deep insights into the Generator and Discriminator modules. This ensures you always have a reference to understand the underlying mechanics.
Explore deeper! For a comprehensive breakdown of the project, its modules, setup, and unique features, please refer to our detailed Wiki. Dive into the specifics of the Generator Architecture and Discriminator Architecture to enhance your understanding and utilization of the project.
- Cloning the Repository
- Setting up the Virtual Environment
- Installing Dependencies
- Configuring the GPU
- Preparing the Dataset
- Configuring Training Parameters
- Training the Model
- Generating Images
- Generating Videos
This project is open-sourced and available to everyone under the MIT License.
For more details check the License and Copyright wiki page.
Contributions are welcome! Feel free to open an issue or submit a pull request if you find any bugs or have suggestions for improvements.
Consult the Contribution and Community Guidelines wiki page.