“…Image generation lungs X-Ray, CT, MRI [2,5,16,17] Image segmentation MRI, CT, ultrasound [9,13,30] Image inpainting MRI [22] Image denoising MRI, CT, retinal OCT [6,11,32] Lesion detection MRI [24,29,31] Image translation MRI, CT [13,15] Seed-image based augmentation Dermatology [23] Skin disease classification Dermatology This work using large synthetic datasets Inspired by the recent early success of DPMs, we propose to use diffusion models for image augmentation as part of supervised machine learning pipelines. More specifically, we study how diffusion models can i) increase the classification metrics for skin diseases, and ii) augment skin condition datasets by effectively manipulating the generated images' features conditioned on the input text prompts.…”