| --- |
| tags: |
| - Lung |
| - Pneumonia |
| - Covid-19 |
| - PyTorch |
| license: mit |
| language: |
| - en |
| pipeline_tag: unconditional-image-generation |
| library_name: diffusers |
| --- |
| --- |
| # Diffusion Model for COVID-19 X-ray Generation |
| This is a diffusion model designed for generating synthetic COVID-19 X-ray images. The model takes random noise as input and iteratively denoises it to produce realistic X-ray images. |
| Used to generate synthetic xray image for scarce COVID-19 positive cases, which can be used for data augmentation in training diagnostic models. |
|
|
| Training data from https://data.mendeley.com/datasets/9xkhgts2s6/4 |
| Full project file at https://github.com/teohyc/covid_xray_diffusion |
|
|
| ##Usage |
| ```python |
| from diffusers import DDPMPipeline |
| import matplotlib.pyplot as plt |
| |
| # Load the pipeline |
| pipeline = DDPMPipeline.from_pretrained("teohyc/Covid-XRay-Diffusion-Model") |
| |
| # Generate a synthetic X-ray |
| image = pipeline(num_inference_steps=500).images[0] #default is 1000 steps, but you can reduce it for faster generation (at the cost of quality) |
| |
| # Display |
| plt.imshow(image, cmap='gray') |
| plt.axis('off') |
| plt.show() |
| ``` |
