Minimum Barrier Salient Object Detection at 80 FPS
We propose a highly efficient, yet powerful, salient object detection method based on the Minimum Barrier Distance (MBD) Transform. The MBD transform is robust to pixel-value fluctuation, and thus can be effectively applied on raw pixels without region abstraction. We present an approximate MBD transform algorithm with 100X speedup over the exact algorithm. An error bound analysis is also provided. Powered by this fast MBD transform algorithm, the proposed salient object detection method runs at 80 FPS, and significantly outperforms previous methods with similar speed on four large benchmark datasets, and achieves comparable or better performance than state-of-the-art methods. Furthermore, a technique based on color whitening is proposed to extend our method to leverage the appearance-based backgroundness cue. This extended version further improves the performance, while still being one order of magnitude faster than all the other leading methods.
PDF Abstract
PASCAL-S
DUT-OMRON
FBMS
SegTrack-v2
FBMS-59
Results from the Paper
Ranked #6 on
Video Salient Object Detection
on VOS-T
(using extra training data)
| Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Uses Extra Training Data |
Benchmark |
|---|---|---|---|---|---|---|---|
| Video Salient Object Detection | DAVIS-2016 | MB+M | S-Measure | 0.600 | # 8 | ||
| MAX E-MEASURE | 0.748 | # 6 | |||||
| AVERAGE MAE | 0.173 | # 3 | |||||
| Video Salient Object Detection | DAVSOD-Difficult20 | MB+M | S-Measure | 0.492 | # 7 | ||
| max E-measure | 0.635 | # 7 | |||||
| Average MAE | 0.251 | # 8 | |||||
| Video Salient Object Detection | DAVSOD-easy35 | MB+M | S-Measure | 0.536 | # 6 | ||
| max E-Measure | 0.624 | # 7 | |||||
| Average MAE | 0.231 | # 8 | |||||
| Video Salient Object Detection | DAVSOD-Normal25 | MB+M | S-Measure | 0.506 | # 6 | ||
| max E-measure | 0.552 | # 8 | |||||
| Average MAE | 0.261 | # 8 | |||||
| Video Salient Object Detection | FBMS-59 | MB+M | S-Measure | 0.609 | # 14 | ||
| AVERAGE MAE | 0.206 | # 15 | |||||
| MAX F-MEASURE | 0.487 | # 14 | |||||
| Video Salient Object Detection | MCL | MB+M | S-Measure | 0.539 | # 8 | ||
| MAX E-MEASURE | 0.733 | # 8 | |||||
| AVERAGE MAE | 0.178 | # 1 | |||||
| Video Salient Object Detection | SegTrack v2 | MB+M | S-Measure | 0.618 | # 8 | ||
| AVERAGE MAE | 0.146 | # 8 | |||||
| max E-measure | 0.778 | # 5 | |||||
| Video Salient Object Detection | UVSD | MB+M | S-Measure | 0.563 | # 6 | ||
| max E-measure | 0.668 | # 7 | |||||
| Average MAE | 0.169 | # 7 | |||||
| Video Salient Object Detection | ViSal | MB+M | S-Measure | 0.726 | # 8 | ||
| max E-measure | 0.832 | # 7 | |||||
| Average MAE | 0.129 | # 8 | |||||
| Video Salient Object Detection | VOS-T | MB+M | S-Measure | 0.661 | # 6 | ||
| max E-measure | 0.698 | # 6 | |||||
| Average MAE | 0.158 | # 7 |
