MASK2TASKS: LEVERAGING SEGMENTATION TO ENHANCE CLASSIFICATION PERFORMANCE IN HISTOPATHOLOGICAL COLORECTAL IMAGES
Keywords: multi-task learning, colorectal hispathological imaging, segmentation, classification
TL;DR: This paper introduces a novel multi-task learning for colorectal hispathological imaging.
Abstract: In this study, we explore the enhancement of colorectal image classification accuracy with the aid of a segmentation task. We introduce Mask2Tasks, a deep neural network for joint colorectal image classification and segmentation which is trained using a novel two-stage training approach. Numerical results have demonstrated its effectiveness in both classification and multi-task learning scenarios.
Supplementary Material: pdf
Submission Number: 124
Loading