Supervised Few-Shot Image Segmentation with Deep Metric Learning
Yanjiang Li · 2021 International Conference on Electronic Information Technology and Smart Agriculture (ICEITSA) · 2021
Conventional deep learning methods have great performance on image segmentation problems with a large number of samples, but directly applying these methods to image segmentation fields where there is a lack of samples is prone to over-fitting, resulting in a significant decline in network segmentation performance. Research on small sample image segmentation methods is of great significance and value for effectively reducing the dependence of these fields on training samples. Among primary small sample image segmentation algorithms, deep metric learning has been widely studied because of its simplicity and efficiency, and its performance mainly depends on feature selection as well as the construction of the network. In order to further improve the segmentation effect of deep metric learning.