Abstract This article proposes a mask refinement method for chromosome instance segmentation. The proposed method exploits the knowledge representation capability of Neural Knowledge DNA (NK-DNA) to capture the semantics of the chromosomes shape, texture, and key points, and then it uses the captured knowledge to improve the accuracy and smoothness of the masks. We validate the methods effectiveness on our latest high-resolution chromosome image dataset. The experimental results show that our proposed methods mask average precision (MaskAP) is 3.66% higher than Mask R-CNN and outperforms advanced Cascade Mask R-CNN by 1.35%.
|