菜根檀的博客

Dense Atrous Spatial Pyramid Pooling-密集空洞空间金字塔池化1 Atrous convolution and pyramid pooling2 Denser feature pyramid and larger receptive field2.1 Denser feature pyramidDenser scale sampling:Denser pixel sampling:2.2 Larger receptive field密集空洞金字塔池化时基于空洞卷积和空洞空

文献:A Simple Pooling-Based Design for Real-Time Salient Object DetectionAbstract1 Introduction2 Related Work3 PoolNet3.1 Overall Pipline3.2 Global Guidance Module3.3 特征聚合模块4 Joint Training with Edge Detection5 ExperimentAbstract    &nb

文献：Learning Depth-Guided Convolutions for Monocular 3D Object​摘要(Abstract)1 介绍(Introduction)2 相关工作(Related Work)3 方法（Methodology）3.1 骨干（Backbone）3.2 深度引导过滤模块（DepthGuidedFiltering Module）---Important细节（details）shift-pooling：shift-convloutional实验结果（Experimen