BTS-Net: Bi-directional Transfer-and-Selection Network For RGB-D Salient Object Detection

5 Apr 2021  ·  Wenbo Zhang, Yao Jiang, Keren Fu, Qijun Zhao ·

Depth information has been proved beneficial in RGB-D salient object detection (SOD). However, depth maps obtained often suffer from low quality and inaccuracy... Most existing RGB-D SOD models have no cross-modal interactions or only have unidirectional interactions from depth to RGB in their encoder stages, which may lead to inaccurate encoder features when facing low quality depth. To address this limitation, we propose to conduct progressive bi-directional interactions as early in the encoder stage, yielding a novel bi-directional transfer-and-selection network named BTS-Net, which adopts a set of bi-directional transfer-and-selection (BTS) modules to purify features during encoding. Based on the resulting robust encoder features, we also design an effective light-weight group decoder to achieve accurate final saliency prediction. Comprehensive experiments on six widely used datasets demonstrate that BTS-Net surpasses 16 latest state-of-the-art approaches in terms of four key metrics. read more

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
RGB-D Salient Object Detection DES BTS-Net S-Measure 94.3 # 1
Average MAE 0.018 # 3
max E-Measure 97.9 # 1
max F-Measure 94.0 # 1
RGB-D Salient Object Detection LFSD BTS-Net S-Measure 86.7 # 2
Average MAE 0.07 # 4
max E-Measure 90.6 # 1
max F-Measure 87.4 # 1
RGB-D Salient Object Detection NJU2K BTS-Net S-Measure 92.1 # 1
Average MAE 0.036 # 3
max E-Measure 95.4 # 1
max F-Measure 92.4 # 1
RGB-D Salient Object Detection NLPR BTS-Net S-Measure 93.4 # 1
Average MAE 0.023 # 3
max F-Measure 92.3 # 2
max E-Measure 96.5 # 1
RGB-D Salient Object Detection SIP BTS-Net S-Measure 89.6 # 1
max E-Measure 93.3 # 2
max F-Measure 90.1 # 1
Average MAE 0.044 # 2
RGB-D Salient Object Detection STERE BTS-Net S-Measure 91.5 # 1
Average MAE 0.038 # 3
max F-Measure 91.1 # 2
max E-Measure 94.9 # 1

Methods


No methods listed for this paper. Add relevant methods here