Attentive Feedback Network for Boundary-Aware Salient Object Detection、
1.Attentive Feedback Network (AFNet) – a Global Perception Module (GPM) Attentive Feedback Modules (AFMs)
2. Boundary-Enhanced Loss (BEL)
Learning Noise-Aware Encoder-Decoder from Noisy Labels by Alternating Back-Propagation for Saliency Detection
1.learn a clean saliency predictor from noisy labels by a novel latent variable model that we call noise-aware encoder-decoder
2. train the latent variable model by an alternating back-propagation(ABP) algorithm
3.an edge-aware smoothness loss
The clean saliency predictive model is an encoder-decoder framework, while the noise is modeled as a non-linear transformation of Gaussian latent variables, in which the transformation is parameterized by a neural network. Edge-aware smoothness loss is also utilized to prevent the model from converging to a trivial solution. We propose to train the model by alternating back-propagation algorithm
Multi-Type Self-Attention Guided Degraded Saliency Detection
- a multi-type self-attention network a multi-type self-attention network(MSA), to select features from both global and local perspectives
- an attention transfer learning network Attention Transfer Network (ATN)-- im-plement recurrent local self-attention (RLS) estimation and global pixel self-attention (GPA) inference.
Multi-spectral Salient Object Detection by Adversarial Domain Adaptation
1.using the synchronized RGB and NIR images for the both simple and complex situations.
2. collect a large multi-spectral dataset
3. the adversarial domain adaptation
Deep Embedding Features for Salient Object Detection
- embedding prior predictions into a metric space to filter outliers and generate attentive features to precisely localize salientobjects.
- Recursive Feature Integration Network(RFIN)
- Guided Filter Refinement Network (GFRN)