DOCK: Detecting Objects by Transferring Common-Sense Knowledge
DOCK: Detecting Objects by Transferring Common-Sense Knowledge
复制标题
DOI：
10.1007/978-3-030-01261-8_30
复制
发表时间：
2018-04
期刊：
影响因子：
--
通讯作者：
Krishna Kumar Singh;S. Divvala;Ali Farhadi;Yong Jae Lee
中科院分区：
文献类型：
--
作者：
Krishna Kumar Singh;S. Divvala;Ali Farhadi;Yong Jae Lee
我们提出了一个可扩展的方法检测对象的常识知识（DOCK）从源到目标类别。在我们的设置中，源类别的训练数据具有边界框注释，而目标类别的训练数据仅具有图像级注释。当前最先进的方法集中在图像级视觉或语义相似性上，以使在源类别上训练的检测器适应新的目标类别。相比之下，我们的关键思想是（i）不在图像级别使用相似性，而是在区域级别使用相似性，以及（ii）利用更丰富的常识（基于属性，空间等）。以引导算法学习正确的检测。我们从现成的知识库中自动获得这些常识性线索，而无需任何额外的人力。在具有挑战性的MS COCO数据集上，我们发现常识知识可以大大提高现有迁移学习基线的检测性能。
We present a scalable approach for Detecting Objects by transferring Common-sense Knowledge (DOCK) from source to target categories. In our setting, the training data for the source categories have bounding box annotations, while those for the target categories only have image-level annotations. Current state-of-the-art approaches focus on image-level visual or semantic similarity to adapt a detector trained on the source categories to the new target categories. In contrast, our key idea is to (i) use similarity not at the image-level, but rather at the region-level, and (ii) leverage richer common-sense (based on attribute, spatial, etc.) to guide the algorithm towards learning the correct detections. We acquire such common-sense cues automatically from readily-available knowledge bases without any extra human effort. On the challenging MS COCO dataset, we find that common-sense knowledge can substantially improve detection performance over existing transfer-learning baselines.