The presence of hair in dermoscopy images poses a challenge to melanoma detection tasks, as it can cause occlusion of crucial diagnostic information. In this paper, we explore the deep-learning approaches intended to address this issue. Specifically, we review notable techniques used for hair removal such as convolutional neural networks, generative models, recurrent neural networks, and autoencoders. We give an overview of these techniques and explore their application, either combined or separately, in the context of hair removal in dermoscopy images. Moreover, we highlight the challenges and discuss future directions.