What Makes for Effective Detection Proposals?
What Makes for Effective Detection Proposals?
复制标题
DOI：
10.1109/tpami.2015.2465908
复制
发表时间：
2016-04-01
影响因子：
23.6
通讯作者：
Schiele, Bernt
中科院分区：
文献类型：
--
作者：
Hosang, Jan;Benenson, Rodrigo;Schiele, Bernt
当前的顶级性能对象检测器采用检测建议来指导对象搜索对象，从而避免了跨图像的详尽滑动窗口搜索。尽管对检测提案的流行和广泛使用，但在对象检测过程中使用它们时，尚不清楚哪种权衡取舍。我们提供了对十二种建议方法的深入分析，以及四个基准，涉及建议可重复性，对Pascal，Imagenet和MS Coco的地面真相注释回忆，以及它们对DPM，R-CNN和快速R-CNN检测性能的影响。我们的分析表明，对于对象检测，提高建议定位的精度与改善召回量同等重要。我们介绍了一个新颖的指标，即平均召回（AR），该指标既奖励高召回率和良好的本地化，又与检测性能非常相关。我们的发现显示了现有方法的共同优势和劣势，并提供了选择和调整建议方法的见解和指标。
Current top performing object detectors employ detection proposals to guide the search for objects, thereby avoiding exhaustive sliding window search across images. Despite the popularity and widespread use of detection proposals, it is unclear which trade-offs are made when using them during object detection. We provide an in-depth analysis of twelve proposal methods along with four baselines regarding proposal repeatability, ground truth annotation recall on PASCAL, ImageNet, and MS COCO, and their impact on DPM, R-CNN, and Fast R-CNN detection performance. Our analysis shows that for object detection improving proposal localisation accuracy is as important as improving recall. We introduce a novel metric, the average recall (AR), which rewards both high recall and good localisation and correlates surprisingly well with detection performance. Our findings show common strengths and weaknesses of existing methods, and provide insights and metrics for selecting and tuning proposal methods.