Cascaded Partial Decoder for Fast and Accurate Salient Object Detection (CVPR2019)

Our model ranks first in the challenging SOC benchmark up to now (2019.11.6).

Requirements:

python2.7, pytorch 0.4.0

Usage

Modify the pathes of backbone and datasets, then run test_CPD.py

Pre-trained model

VGG16 backbone: google drive, BaiduYun (code: gb5u)

ResNet50 backbone: google drive, BaiduYun (code: klfd)

Pre-computed saliency maps

VGG16 backbone: google drive

ResNet50 backbone: google drive

Performance

Maximum F-measure

Model	FPS	ECSSD	HKU-IS	DUT-OMRON	DUTS-TEST	PASCAL-S
PiCANet	7	0.931	0.921	0.794	0.851	0.862
CPD	66	0.936	0.924	0.794	0.864	0.866
PiCANet-R	5	0.935	0.919	0.803	0.860	0.863
CPD-R	62	0.939	0.925	0.797	0.865	0.864

MAE

Model	ECSSD	HKU-IS	DUT-OMRON	DUTS-TEST	PASCAL-S
PiCANet	0.046	0.042	0.068	0.054	0.076
CPD	0.040	0.033	0.057	0.043	0.074
PiCANet-R	0.046	0.043	0.065	0.051	0.075
CPD-R	0.037	0.034	0.056	0.043	0.072

Shadow Detection

pre-computed maps: google drive

Performance

BER

Model	SBU	ISTD	UCF
DSC	5.59	8.24	8.10
CPD	4.19	6.76	7.21

Citation

@InProceedings{Wu_2019_CVPR,
author = {Wu, Zhe and Su, Li and Huang, Qingming},
title = {Cascaded Partial Decoder for Fast and Accurate Salient Object Detection},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2019}
}