This project is an implementation of Deep Convolutional Generative Adversarial Networks (DCGANs) using PyTorch, structured with Object-Oriented Programming (OOP) principles.

• Operating System: Windows 11
• GPU: Nvidia GTX 1050 Ti
• CPU: Intel Core i7-8750H
• RAM: 16 GB
• CUDA Version: 11.8 (installed via PyTorch using pip command)
• Anaconda: Python 3.11+

Before running the project, ensure that Nvidia CUDA is installed on your PC.

• CHANNELS_IMG:
  • Automatically uncommented: Assumes the dataset is RGB images with 3 channels.
  • Uncomment manually: If using MNIST dataset or any dataset with monocolor (1 channel).
• Z_DIM: Noise dimension (set to 100).
• NUM_EPOCHS: Number of training epochs (set to 50).

To use the MNIST dataset:

• Uncomment line 25 in Training.py.
• Comment line 24.

Note: The MNIST dataset is available to download via a Python script. Other datasets must be downloaded manually and placed in their respective folders (/MRI/, /celeb_dataset/, etc.).

• For 3 channels: Uncomment line 19 and comment line 20 in Training.py.
• For 1 channel: Uncomment line 20 and comment line 19 in Training.py.

transform = transforms.Compose([
    transforms.Resize((IMAGE_SIZE, IMAGE_SIZE)),
    transforms.ToTensor(),
    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])

transform = transforms.Compose([
    transforms.Resize((IMAGE_SIZE, IMAGE_SIZE)),
    transforms.ToTensor(),
    # transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
    transforms.Normalize((0.5,), (0.5,))
])

Here are two examples of generated images after training:

You can test the project by running the Test.py file with the following command:

python Test.py

or

python main.py

To run the script:

python Training.py