salient object detection

A Simple Pooling-Based Design for Real-Time Salient Object Detection （2019 CVPR）

Global Guidance Module (GGM) : PPM feature 与 FPN 每个level 融合。增加RF，缓解obj预测不完整。
Feature Aggregation Module (FAM): 类似PPM，add代替concate。
       1、reducing the aliasing effect of upsampling。
       2、allows each spatial location to view the local context at different scale spaces, context at different scale spaces
joint training：edge data（BCE）和 salient data （CE）交替训练，细化edge。

Contrast Prior and Fluid Pyramid Integration for RGBD Salient Object Detection （2019 CVPR）

depth label 与 SOD 结合。
Contrast-enhanced Net(CEN)：增加depth map前景背景的contrast，类内紧凑，类间可分。
original depth map contrast not obvious enough
    
cross-modal fusion：Enhanced map 作为 attention map  
Fluid Pyramid Integration(FPI)：FPI leads all the highlevel features into low-level features for every node.
B+C>B+C+M, meets feature compatibility problem
LOSS：saliency cross-entropy loss + 5 contrast loss