Cascaded Partial Decoder for Fast and Accurate Salient Object Detection

CVPR 2019  ·  Zhe Wu, Li Su, Qingming Huang ·

Existing state-of-the-art salient object detection networks rely on aggregating multi-level features of pre-trained convolutional neural networks (CNNs). Compared to high-level features, low-level features contribute less to performance but cost more computations because of their larger spatial resolutions... In this paper, we propose a novel Cascaded Partial Decoder (CPD) framework for fast and accurate salient object detection. On the one hand, the framework constructs partial decoder which discards larger resolution features of shallower layers for acceleration. On the other hand, we observe that integrating features of deeper layers obtain relatively precise saliency map. Therefore we directly utilize generated saliency map to refine the features of backbone network. This strategy efficiently suppresses distractors in the features and significantly improves their representation ability. Experiments conducted on five benchmark datasets exhibit that the proposed model not only achieves state-of-the-art performance but also runs much faster than existing models. Besides, the proposed framework is further applied to improve existing multi-level feature aggregation models and significantly improve their efficiency and accuracy. read more

PDF Abstract CVPR 2019 PDF CVPR 2019 Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Camouflaged Object Segmentation COD CPD MAE 0.059 # 4
Weighted F-Measure 50.8 # 4
S-Measure 74.7 # 3
E-Measure 77.0 # 4
RGB Salient Object Detection DUT-OMRON CPD-R (ResNet50) MAE 0.056 # 4
F-measure 0.747 # 3
RGB Salient Object Detection DUTS-test CPD-R (ResNet50) F-measure 80.5 # 1
MAE 0.043 # 2
RGB Salient Object Detection ECSSD CPD-R (ResNet50) MAE 0.037 # 2
F-measure 0.917 # 3
RGB Salient Object Detection HKU-IS CPD-R (ResNet50) MAE 0.034 # 5
F-measure 0.891 # 3
RGB Salient Object Detection ISTD CPD Balanced Error Rate 6.76 # 1
RGB Salient Object Detection PASCAL-S CPD-R (ResNet50) MAE 0.072 # 2
F-measure 0.824 # 3
RGB Salient Object Detection SBU CPD Balanced Error Rate 4.19 # 1
RGB Salient Object Detection UCF CPD Balanced Error Rate 7.21 # 1

Methods