PyTorchGANsDiffusion ModelsMedical AIComputer Vision
Synthetic Pneumonia Image Generation Research
September 2024 – December 2024
A comprehensive research project exploring the use of generative models for medical image augmentation. The research compared GANs and diffusion models for generating synthetic chest X-rays to augment limited medical datasets. The diffusion model achieved FID 2.54 compared to GAN's 11.26, demonstrating superior image quality.
Key Highlights
Built pipelines for synthetic chest X-ray generation
Achieved 87.02% test accuracy using GAN augmentation
Diffusion models achieved FID 2.54 vs GAN 11.26
Performed Grad-CAM explainability analysis
Authored peer-reviewed research paper
Tech Stack
ML Framework
PyTorchtorchvisionNumPy
Models
StyleGAN2DDPMResNet
Analysis
Grad-CAMFID Scorescikit-learn
Infrastructure
CUDAWeights & BiasesJupyter
Interested in this project?
Let's discuss how we can collaborate or learn more about the implementation details.