Corpus ID: 235458061

Dynamic Knowledge Distillation with A Single Stream Structure for RGB-D Salient Object Detection

  title={Dynamic Knowledge Distillation with A Single Stream Structure for RGB-D Salient Object Detection},
  author={Guangyu Ren and T. Stathaki},
RGB-D salient object detection(SOD) demonstrates its superiority on detecting in complex environments due to the additional depth information introduced in the data. Inevitably, an independent stream is introduced to extract features from depth images, leading to extra computation and parameters. This methodology which sacrifices the model size to improve the detection accuracy may impede the practical application of SOD problems. To tackle this dilemma, we propose a dynamic distillation method… Expand

Figures and Tables from this paper


A2dele: Adaptive and Attentive Depth Distiller for Efficient RGB-D Salient Object Detection
A depth distiller (A2dele) is proposed to explore the way of using network prediction and attention as two bridges to transfer the depth knowledge from the depth stream to the RGB stream and achieves state-of-the-art performance. Expand
Progressively Guided Alternate Refinement Network for RGB-D Salient Object Detection
This paper proposes a progressively guided alternate refinement network for RGB-D salient object detection, which outperforms existing state-of-the-art approaches by a large margin, and shows superiority in efficiency and model size. Expand
Hierarchical Dynamic Filtering Network for RGB-D Salient Object Detection
A hybrid enhanced loss function is designed to make the predictions have sharper edges and consistent saliency regions, and implement a kind of more flexible and efficient multi-scale cross-modal feature processing, i.e. dynamic dilated pyramid module. Expand
Contrast Prior and Fluid Pyramid Integration for RGBD Salient Object Detection
Contrast prior is utilized, which used to be a dominant cue in none deep learning based SOD approaches, into CNNs-based architecture to enhance the depth information and is integrated with RGB features for SOD, using a novel fluid pyramid integration. Expand
Rethinking RGB-D Salient Object Detection: Models, Data Sets, and Large-Scale Benchmarks
It is demonstrated that D3Net can be used to efficiently extract salient object masks from real scenes, enabling effective background-changing application with a speed of 65 frames/s on a single GPU. Expand
Adaptive Fusion for RGB-D Salient Object Detection
Two-streamed convolutional neural networks are designed, each of which extracts features and predicts a saliency map from either RGB or depth modality, and this approach outperforms state-of-the-art methods on three publicly available datasets. Expand
CNNs-Based RGB-D Saliency Detection via Cross-View Transfer and Multiview Fusion.
A novel framework based on convolutional neural networks (CNNs), which transfers the structure of the RGB-based deep neural network to be applicable for depth view and fuses the deep representations of both views automatically to obtain the final saliency map is proposed. Expand
RGBD Salient Object Detection via Deep Fusion
A new convolutional neural network is designed to automatically learn the interaction mechanism for RGBD salient object detection by adopting various flexible and interpretable saliency feature vectors as inputs and integrates a superpixel-based Laplacian propagation framework with the trained CNN to extract a spatially consistent saliency map. Expand
RGBD Salient Object Detection: A Benchmark and Algorithms
A simple fusion framework that combines existing RGB-produced saliency with new depth-induced saliency and a specialized multi-stage RGBD model is proposed which takes account of both depth and appearance cues derived from low-level feature contrast, mid-level region grouping and high-level priors enhancement. Expand
Three-Stream Attention-Aware Network for RGB-D Salient Object Detection
  • H. Chen, Youfu Li
  • Computer Science, Medicine
  • IEEE Transactions on Image Processing
  • 2019
In the proposed architecture, a cross-modal distillation stream, accompanying the RGB-specific and depth-specific streams, is introduced to extract new RGB-D features in each level in the bottom–up path, and the channel-wise attention mechanism is innovatively introduced to the cross- modal cross-level fusion problem to adaptively select complementary feature maps from each modality in eachlevel. Expand