ProgressGAN-CXR is a collection of four severity-conditioned GAN generators trained on the COVID-19 Radiography Database for chest X-ray disease progression simulation. Given a continuous severity score from 0.0 (Normal) to 1.0 (COVID-19), the generator produces a synthetic chest X-ray representing that disease stage. The repository contains four model checkpoints: DCGAN (best FID: 142.14), WGAN-GP (best semantic encoding among baselines: Spearman r 0.748), Spectral DCGAN (best disentanglement: Spearman r 0.996), and KD Generator (best overall: FID 143.17, Spearman r 0.984, SSIM 0.963). Also includes a ResNet18 classifier fine-tuned for four-class COVID-19 CXR classification at 95.28% accuracy. Intended for medical education tooling, data augmentation research, and GAN evaluation methodology.
