Author(s)

Chen Mingyang

Corresponding Author

Chen Mingyang

Abstract

In recent years, Mask-R-CNN has achieved great success in the field of image segmentation. This paper proposes an accurate image segmentation method based on Mask-R-CNN.We improve the boundary segmentation precision by modifying the mask branch structure. Used feature fusion on the mask branch to enhance the combination of semantic information and shallow information, and improve the segmentation effect. Expanded the size of the feature map of the RoIAlign layer, get more accurate location information. The experiment shows the accuracy of the algorithm.

Keywords

Segmentation Algorithm, Improvement