BASNet: Boundary-Aware Salient Object Detection

Deep Convolutional Neural Networks have been adopted for salient object detection and achieved the state-of-the-art performance. Most of the previous works however focus on region accuracy but not on the boundary quality... In this paper, we propose a predict-refine architecture, BASNet, and a new hybrid loss for Boundary-Aware Salient object detection. Specifically, the architecture is composed of a densely supervised Encoder-Decoder network and a residual refinement module, which are respectively in charge of saliency prediction and saliency map refinement. The hybrid loss guides the network to learn the transformation between the input image and the ground truth in a three-level hierarchy -- pixel-, patch- and map- level -- by fusing Binary Cross Entropy (BCE), Structural SIMilarity (SSIM) and Intersection-over-Union (IoU) losses. Equipped with the hybrid loss, the proposed predict-refine architecture is able to effectively segment the salient object regions and accurately predict the fine structures with clear boundaries. Experimental results on six public datasets show that our method outperforms the state-of-the-art methods both in terms of regional and boundary evaluation measures. Our method runs at over 25 fps on a single GPU. The code is available at: https://github.com/NathanUA/BASNet. read more

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Benchmark
Camouflaged Object Segmentation CAMO BASNet MAE 0.159 # 9
Weighted F-Measure 41.3 # 9
S-Measure 61.8 # 9
E-Measure 66.1 # 9
Camouflaged Object Segmentation COD BASNet MAE 0.105 # 6
Weighted F-Measure 36.5 # 5
S-Measure 63.4 # 6
E-Measure 67.8 # 6
RGB Salient Object Detection DUT-OMRON BASNet MAE 0.056 # 4
Salient Object Detection DUTS-TE BASNET (ResNet-34) {max}Fβ 0.86 # 1
MAE 0.047 # 2
Fwβ 0.803 # 1
Sm 0.853 # 1
relaxFbβ 0.758 # 1
RGB Salient Object Detection DUTS-TE BASNet MAE 0.047 # 6
S-Measure 0.876 # 3
mean E-Measure 0.896 # 3
mean F-Measure 0.823 # 3
Salient Object Detection ECSSD BASNet (ResNet-34) Model Size (MB) 348.5 # 1
{max}Fβ 0.942 # 2
MAE 0.037 # 2
Fwβ 0.904 # 1
Sm 0.916 # 2
relaxFbβ 0.826 # 1
RGB Salient Object Detection ECSSD BASNet MAE 0.037 # 2
RGB Salient Object Detection HKU-IS BASNet MAE 0.032 # 4
RGB Salient Object Detection PASCAL-S BASNet MAE 0.076 # 4
RGB Salient Object Detection SOC BASNet S-Measure 0.841 # 3
mean E-Measure 0.864 # 3
Average MAE 0.092 # 3
RGB Salient Object Detection SOD BASNet MAE 0.114 # 3

Methods


No methods listed for this paper. Add relevant methods here