| 内容提要: |
Semantic segmentation can be used to help improve object detection. However, semantic segmentation usually rely on accurate and pixel-wise ground truth segmentation mask, which is very expensive to obtain. Meanwhile, most works only jointly learns semantic segmentation and objection detection independently. In this report, we introduce a pseudo mask augmented object detection, which generate object mask based on object bounding box.
[1]Xiangyun Zhao, Shuang Liang, Yichen Wei, Pseudo Mask Augmented Object Detection, CoRR abs/1803.05858, 2018.
|