Scholarlink

Target detection tasks need to incorporate scene semantic understanding in order to achieve more precise instance segmentation. Instance segmentation is a challenging research task but is fundamental in many applications. In this paper, we propose an Attention Fused Mask R-CNN Mechanism (AF Mask R-CNN), which is based on a dual attention mechanism, and is designed for detecting and segmenting instances in visible images making use of semantic information. The experimental results obtained using the Coco 2017 dataset show that our AF Mask R-CNN model can effectively improve the segmentation accuracy. We are able to demonstrate that AF Mask R-CNN outperforms current instance segmentation deep learning algorithms.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AF Mask R-CNN: Attention Fused Mask R-CNN Mechanism for Instance Segmentation

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Zhichao Su,
  • Chengcai Leng,
  • Irene Cheng

摘要

Target detection tasks need to incorporate scene semantic understanding in order to achieve more precise instance segmentation. Instance segmentation is a challenging research task but is fundamental in many applications. In this paper, we propose an Attention Fused Mask R-CNN Mechanism (AF Mask R-CNN), which is based on a dual attention mechanism, and is designed for detecting and segmenting instances in visible images making use of semantic information. The experimental results obtained using the Coco 2017 dataset show that our AF Mask R-CNN model can effectively improve the segmentation accuracy. We are able to demonstrate that AF Mask R-CNN outperforms current instance segmentation deep learning algorithms.