arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2005.02699cs.CVcs.LGeess.IV
收藏

ProbaNet:用于目标检测的建议平衡网络

ProbaNet: Proposal-balanced Network for Object Detection

University of Electronic Science and Technology of China(电子科技大学)

Jing Wu, Xiang Zhang, Mingyi Zhou, Ce Zhu

摘要:

基于卷积神经网络(CNN)的目标检测器产生的候选目标建议会遇到易-难样本不平衡的问题,这会影响整体性能。在这项研究中,我们提出了一个建议平衡网络(ProbaNet),以减轻不平衡的问题。首先,ProbaNet通过阈值截断丢弃容易的样本来增加选择硬样本进行训练的概率。其次,ProbaNet通过增加权重来强调前景提案。为了评估ProbaNet的有效性,我们基于不同的基准对模型进行训练。使用ProbaNet的模型的平均精度(mAP)比PASCAL VOC 2007的基线高1.2 $\%$。此外,它与现有的两级检测器兼容,并提供了非常少量的额外计算成本。

英文摘要:

Candidate object proposals generated by object detectors based on convolutional neural network (CNN) encounter easy-hard samples imbalance problem, which can affect overall performance. In this study, we propose a Proposal-balanced Network (ProbaNet) for alleviating the imbalance problem. Firstly, ProbaNet increases the probability of choosing hard samples for training by discarding easy samples through threshold truncation. Secondly, ProbaNet emphasizes foreground proposals by increasing their weights. To evaluate the effectiveness of ProbaNet, we train models based on different benchmarks. Mean Average Precision (mAP) of the model using ProbaNet achieves 1.2$\%$ higher than the baseline on PASCAL VOC 2007. Furthermore, it is compatible with existing two-stage detectors and offers a very small amount of additional computational cost.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑