Cascaded Partial Decoder for Fast and Accurate Salient Object Detection
1.We propose a novel cascaded partial decoder frame-work, which discards low-level features to reduce the complexity of deep aggregation models, and utilizes generated relatively precise attention map to refine high-level features to improve the performance.
Multi-source weak supervision for saliency detection
1.We propose a novel weak supervision framework to train saliency detection models with diverse supervision source:
a multi-label classification network (CNet), a caption gen-eration network (PNet) and a saliency prediction network
(SNet).
2. An attention coherence loss is defined on unlabelled data to encourage the networks to detect gen-erally salient regions instead of task-specific regions. We use CNet and PNet to generate pixel-level pseudo labels to train a saliency prediction network (SNet).
A Mutual Learning Method for Salient Object Detection with intertwined Multi-Supervision
1.using multi-task intertwined supervision,
2.the foreground contour detection and edge detection task to guide each other
An Iterative and Cooperative Top-down and Bottom-up Inference Network for Salient Object Detection
- imitating top-down and bottom-up human perception processes
2.frame-work to joint and iterative way
know: VGG16net and resnet50 --CNN layer
convRNN, convLSTM, convGRU – how to use
Pyramid Feature Attention Network for Saliency detection
- Context-aware Pyramid Feature Extraction (CPFE) module —capture rich context features—get multi-scale
multi-receptive-field high-level features
2.channel-wise attention (CA) after CPFE feature maps and spatial attention (SA) after low-level feature maps
2019 cvpr