Skip to content

Medical Imaging Datasets

Datasets for training diffusion models and other generative models on medical images.

Available Datasets

Dataset Type Size Resolution Document
Synthetic X-Ray Generated Configurable Configurable chest_xray.md
Kaggle Chest X-Ray Real ~5,800 Variable chest_xray.md

Use Cases

Medical imaging datasets in genai-lab are used for:

  1. Diffusion model development: Testing score networks, noise schedules, sampling
  2. DiT training: Demonstrating Transformer-based diffusion
  3. Flow matching: Rectified flow on realistic image data
  4. Benchmarking: Comparing architectures on structured data
Module Purpose
src/genailab/diffusion/datasets.py SyntheticXRayDataset, ChestXRayDataset
src/genailab/diffusion/architectures.py UNet2D, UNet3D for image diffusion
src/genailab/diffusion/training.py train_image_diffusion
  • notebooks/diffusion/03_medical_imaging_diffusion/ — Full pipeline demo

Documents