GANs for Medical Imaging: Synthesizing Realistic Images for Analysis
College Project
Summary
This project focused on generating synthetic medical images using DCGAN, a type of Generative Adversarial Network specialized in image generation. The primary goal was to address data scarcity in the medical domain by producing realistic medical images that could augment existing datasets. The model was trained on limited medical image data, which posed challenges like slow training, low-quality outputs in early epochs, and frequent timeouts due to resource constraints. Over time, image clarity and realism improved with more training epochs. Despite hardware and data limitations, the project demonstrated the potential of GANs in medical image synthesis for research and augmentation purposes.