This is a repository for a micro project on Generative Adversarial Networks (GANs) for MINST dataset implemented using PyTorch. GANs are a class of deep learning models used for generative tasks such as image synthesis, text generation, and video generation. In this project we will try to generate fake MINST images using Generator - Discriminator Architecture of GANs.
The project involves utilizing Generative Adversarial Networks (GANs) to generate fake MNIST images. GANs are a class of machine learning models consisting of two components: a generator and a discriminator. The generator learns to create realistic data (in this case, images) that resemble the MNIST dataset, while the discriminator learns to distinguish between real and fake images.
The project will likely involve the following steps:
-
Preparing and understanding the MNIST dataset: The MNIST dataset is a collection of handwritten digit images, commonly used in machine learning. The project will involve studying and preprocessing the dataset to ensure it is suitable for training the GAN.
-
Designing and training the GAN model: This step involves designing and implementing the generator and discriminator components of the GAN model. The generator will learn to generate fake MNIST images, while the discriminator will learn to distinguish between real and fake images. The model will then be trained using the MNIST dataset.
-
Evaluating the GAN's performance: After training the model, its performance will be evaluated. This may involve assessing how well the generated images resemble the original MNIST dataset. Metrics such as visual inspection, accuracy, or other image quality metrics can be used to evaluate the output.
-
Fine-tuning and improving the GAN: If the initial results are not satisfactory, the GAN can be fine-tuned and improved. This may involve adjusting hyperparameters, adding regularizations, or introducing architectural changes to the model.
-
Generating fake MNIST images: Once the GAN model is trained and deemed satisfactory, it can be used to generate fake MNIST images. These generated images may closely resemble the original MNIST dataset, despite not being real.
Overall, this project aims to demonstrate the capability of GANs to generate convincing fake MNIST images, showcasing the power and potential of generative models in the field of computer vision.
-
Clone the repository:
git clone https://github.com/tajammulbasheer/GAN-for-MNIST.git -
Navigate to the project directory:
cd GAN-for-MNIST -
Create a new virtual environment:
python -m venv env -
Activate the virtual environment:
-
On Windows:
env\Scripts\activate -
On Linux or macOS:
source env/bin/activate
-
-
Install the required libraries:
pip install -r requirements.txtThis will install the following libraries:
- PyTorch
- scikit-learn
- NumPy
- Pandas
- Matplotlib
This project was created for understanding and learning PyTorch for GAN for MNIST Images. There nothing hard to understand in the project as we are only creating a basic GAN. To use the project, follow these steps:
-
Clone the repository:
git clone https://github.com/tajammulbasheer/GAN-for-MNIST.git -
Navigate to the project directory:
cd GAN-for-MNIST -
Install the required libraries:
pip install -r requirements.txt -
Explore the project notebook:
The project include a Jupyter notebook that demonstrate how to create a GAN for MNIST. To run the notebooks, use the following command:
jupyter notebook
I would like to thank the creators of the MNIST datset for making it available for research and educational purposes. We would also like to thank the developers of PyTorch,scikit-learn, NumPy, Pandas, Matplotlib, Seaborn, and OpenCV for their contributions to open-source software, which made this mini- project possible.
If you have any questions or feedback, please feel free to contact me at email. We would be happy to hear from you