论文笔记之：Instance-aware Semantic Segmentation via Multi-task Network Cascades

Instance-aware Semantic Segmentation via Multi-task Network Cascades

Jifeng Dai Kaiming He Jian Sun

本文的出发点是做Instance-aware Semantic Segmentation，但是为了做好这个，作者将其分为三个子任务来做：

1) Differentiating instances. 实例区分

2) Estimating masks.　　　　掩膜估计

3) Categorizing objects.　　分类目标

通过这种分解，作者提出了如下的多任务学习框架，即：Multi-task Network Cascades (MNCs)，示意流程如下：

下面详细的介绍下这个流程，即：

1. Multi-task Network Cascades

1). Regressing Box-level Instances

　　第一个阶段是回归出物体的bbox，这是一个全卷积的子网络。本文follow了Faster R-CNN的提取proposal的方法Region Proposal Networks (RPNs)。在共享feature之前，作者先用了一个 3*3的Conv 用于降维，紧跟着用2个1*1的Conv层回归出其位置，并且对目标进行分类。该阶段的loss function是：

其中，B是该阶段的输出，是一系列的box，B = { Bi }, Bi = { xi; yi; wi; hi; pi }，box的中心点和长宽分别是：xi yi wi hi, yi是物体的概率。

2). Regressing Mask-level Instances

　　该阶段的输出是对每一个box的proposal进行像素级的mask分割。

　　 Given a box predicted by stage 1, we extract a feature of this box by Region-of-Interest (RoI) pooling . The purpose of RoI pooling is for producing a fixed-size feature from an arbitrary box, which is set as 14*

论文笔记之：Instance-aware Semantic Segmentation via Multi-task Network Cascades相关推荐

论文笔记-Domain Adaptation for Semantic Segmentation with Maximum Squares Loss
论文信息论文标题:Domain Adaptation for Semantic Segmentation with Maximum Squares Loss 论文作者:Minghao Chen, H ...
论文笔记：STD2P: RGBD Semantic Segmentation Using Spatio-Temporal Data-Driven Pooling
STD2P: RGBD Semantic Segmentation Using Spatio-Temporal Data-Driven Pooling Yang He, Wei-Chen Chiu, ...
论文笔记-Understanding Convolution for Semantic Segmentation
图森和CMU的合作工作. 论文链接[https://arxiv.org/abs/1702.08502](https://arxiv.org/abs/1702.08502) 主要提出DUC(dense ...
JSNet: Joint Instance and Semantic Segmentation of 3D Point Clouds
Abstract 在本文中,我们提出了一种新的联合实例和语义分割方法,称为JSNet,以同时解决3D点云的实例和语义分割.首先,我们建立了一个有效的骨干网络来从原始点云中提取鲁棒的特征.其次,为了获得 ...
ECCV2020语义分割——Self-Prediction for Joint Instance and Semantic Segmentation of Point Clouds
Self-Prediction for Joint Instance and Semantic Segmentation of Point Clouds Abstract (一) Introducti ...
论文笔记 Object-Aware Instance Labeling for Weakly Supervised Object Detection - ICCV 2019
Object-Aware Instance Labeling for Weakly Supervised Object Detection Kosugi ICCV, 2019 (PDF) (Citat ...
【论文阅读】Weakly Supervised Semantic Segmentation using Out-of-Distribution Data
一篇弱监督分割领域的论文,发表在CVPR2022上: 论文标题: Weakly Supervised Semantic Segmentation using Out-of-Distribution D ...
语义分割--Large Kernel Matters--Improve Semantic Segmentation by Global Convolutional Network
Large Kernel Matters–Improve Semantic Segmentation by Global Convolutional Network https://arxiv.org ...
Robust semantic segmentation by dense fusion network on blurred vhr remote sensing images
Robust semantic segmentation by dense fusion network on blurred vhr remote sensing images 说明se的设计还是很 ...

论文笔记之：Instance-aware Semantic Segmentation via Multi-task Network Cascades

论文笔记之：Instance-aware Semantic Segmentation via Multi-task Network Cascades相关推荐

最新文章

热门文章