Title:Efficient Palm-Line Segmentation with U-Net Context Fusion Module

Abstract:Many cultures around the world believe that palm reading can be used to predict the future life of a person. Palmistry uses features of the hand such as palm lines, hand shape, or fingertip position. However, the research on palm-line detection is still scarce, many of them applied traditional image processing techniques. In most real-world scenarios, images usually are not in well-conditioned, causing these methods to severely under-perform. In this paper, we propose an algorithm to extract principle palm lines from an image of a person's hand. Our method applies deep learning networks (DNNs) to improve performance. Another challenge of this problem is the lack of training data. To deal with this issue, we handcrafted a dataset from scratch. From this dataset, we compare the performance of readily available methods with ours. Furthermore, based on the UNet segmentation neural network architecture and the knowledge of attention mechanism, we propose a highly efficient architecture to detect palm-lines. We proposed the Context Fusion Module to capture the most important context feature, which aims to improve segmentation accuracy. The experimental results show that it outperforms the other methods with the highest F1 Score about 99.42% and mIoU is 0.584 for the same dataset.