Skin lesion segmentation using U-Net and good training strategies
This work addresses skin lesion segmentation for melanoma detection, but it is incremental as it builds on the established U-Net architecture with training optimizations.
The paper tackled skin lesion segmentation by applying a U-Net-based convolutional neural network with improved training strategies, achieving a threshold Jaccard index of 77.5% on the ISIC Challenge 2018 dataset.
In this paper we approach the problem of skin lesion segmentation using a convolutional neural network based on the U-Net architecture. We present a set of training strategies that had a significant impact on the performance of this model. We evaluated this method on the ISIC Challenge 2018 - Skin Lesion Analysis Towards Melanoma Detection, obtaining threshold Jaccard index of 77.5%.