Asymmetric Two-Stream Architecture for Accurate RGB-D Saliency Detection学习笔记

最新推荐文章于 2022-07-12 16:10:45 发布

木子木木峰

最新推荐文章于 2022-07-12 16:10:45 发布

阅读量436

点赞数 1

分类专栏：论文文章标签：计算机视觉深度学习

本文链接：https://blog.csdn.net/choucdaf/article/details/108753764

版权

该博客探讨了在复杂场景中RGB图像特征的局限性，提出了一种不对称双流架构，通过Flow Ladder Module (FLM)、DepthNet和Depth Attention Module (DAM)来保留局部细节和提取全局信息。DepthNet利用深度图获取空间细节，RGBnet结合VGG19和FEM捕捉结构信息。DAM利用深度线索生成注意力权重，精确定位显著目标。

摘要由CSDN通过智能技术生成

一、背景

**
1.对某些复杂情况的RGB图像特征难以鉴别。

appearance features in RGB data are less predictive to some challenging scenes

2.由于RGB图像的复杂特征使用对称两步流模型会“overlooked”，从而产生大量损失。

a symmetric two-stream network may overlook the inherent differences of RGB and depth data.

existing RGB-D methods inevitably suffer from detail information loss [41,16] for adopting strides and pooling operations in the RGB and depth streams.

3.现有的解决方案限制预测特征结构和细节。

An intuitive solution is to use skip-connections [22] or short-connections [21] for reconstructing the detail information.Although these strategies have brought satisfactory improvements, they remain restrictive to predict the complete structures with fine details.