Simple U-net based synthetic polyp image generation: Polyp to negative and negative to polyp

Synthetic polyp generation is a good alternative to overcome the privacy problem of medical data and lack various samples. In this study, we propose deep learning-based image framework that generates synthetic images are similar real ones. We suggest converts given into negative (image without polyp) using simple conditional GAN architecture then new-looking same network. addition, by controllable masks, polyps with characteristics can be generated from one input condition. The used directly as training for detection segmentation additional labeling. To quantitatively assess quality polyps, use public video datasets combined examine performance improvement several models. Experimental results show obtain gains when added set.