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.
PDF AbstractCode
Colab
PASCAL-S
DUTS
HKU-IS
DUT-OMRON
COD10K
CAMO
SOC
DIS5K
ECSSD
| Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Uses Extra Training Data |
Benchmark |
|---|---|---|---|---|---|---|---|
| Camouflaged Object Segmentation | CAMO | BASNet | MAE | 0.159 | # 13 | ||
| Weighted F-Measure | 0.413 | # 13 | |||||
| S-Measure | 0.618 | # 12 | |||||
| Camouflaged Object Segmentation | COD | BASNet | MAE | 0.092 | # 10 | ||
| Weighted F-Measure | 0.352 | # 10 | |||||
| S-Measure | 0.685 | # 10 | |||||
| Dichotomous Image Segmentation | DIS-TE1 | BASNet | max F-Measure | 0.688 | # 8 | ||
| weighted F-measure | 0.595 | # 7 | |||||
| MAE | 0.084 | # 8 | |||||
| S-Measure | 0.754 | # 8 | |||||
| E-measure | 0.801 | # 7 | |||||
| HCE | 220 | # 6 | |||||
| Dichotomous Image Segmentation | DIS-TE2 | BASNet | max F-Measure | 0.755 | # 8 | ||
| weighted F-measure | 0.668 | # 6 | |||||
| MAE | 0.084 | # 6 | |||||
| S-Measure | 0.786 | # 8 | |||||
| E-measure | 0.836 | # 7 | |||||
| HCE | 480 | # 7 | |||||
| Dichotomous Image Segmentation | DIS-TE3 | BASNet | max F-Measure | 0.785 | # 8 | ||
| weighted F-measure | 0.696 | # 10 | |||||
| MAE | 0.083 | # 10 | |||||
| S-Measure | 0.798 | # 9 | |||||
| E-measure | 0.856 | # 10 | |||||
| HCE | 948 | # 7 | |||||
| Dichotomous Image Segmentation | DIS-TE4 | BASNet | max F-Measure | 0.780 | # 8 | ||
| weighted F-measure | 0.693 | # 7 | |||||
| MAE | 0.091 | # 7 | |||||
| S-Measure | 0.794 | # 8 | |||||
| E-measure | 0.848 | # 7 | |||||
| HCE | 3601 | # 7 | |||||
| Dichotomous Image Segmentation | DIS-VD | BASNet | max F-Measure | 0.731 | # 8 | ||
| weighted F-measure | 0.641 | # 7 | |||||
| MAE | 0.094 | # 7 | |||||
| S-Measure | 0.768 | # 8 | |||||
| E-measure | 0.816 | # 9 | |||||
| HCE | 1402 | # 7 | |||||
| RGB Salient Object Detection | DUT-OMRON | BASNet | MAE | 0.056 | # 12 | ||
| RGB Salient Object Detection | DUTS-TE | BASNet | MAE | 0.047 | # 16 | ||
| S-Measure | 0.876 | # 11 | |||||
| mean E-Measure | 0.896 | # 7 | |||||
| mean F-Measure | 0.823 | # 7 | |||||
| RGB Salient Object Detection | ECSSD | BASNet | MAE | 0.037 | # 10 | ||
| RGB Salient Object Detection | HKU-IS | BASNet | MAE | 0.032 | # 11 | ||
| RGB Salient Object Detection | PASCAL-S | BASNet | MAE | 0.076 | # 11 | ||
| 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 |
