Hair removal in dermoscopy images using variational autoencoders

Abstract

The evaluation of the hair-free reconstructed images is carried out using t-distributed stochastic neighbor embedding (SNE) feature mapping by visualizing the distribution of the real hair-free images and the synthesized hair-free images. The conducted experiments on the publicly available dataset HAM10000 show that our method is very efficient.