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 2009.01972cs.CV
收藏

基于统计信息的属性自适应容限Softmax损失

Attribute Adaptive Margin Softmax Loss using Privileged Information

West Virginia University(西弗吉尼亚大学)

Seyed Mehdi Iranmanesh, Ali Dabouei, Nasser M. Nasrabadi

摘要:

我们提出了一个新的框架,利用特权信息识别,这是只在训练阶段提供。在这里,我们专注于识别任务,其中图像作为主视图提供,软生物特征(属性)作为特权数据(仅在训练阶段可用)提供。我们证明,通过实施深度网络来调整利用属性的类之间的自适应边缘,可以学习更具区分力的特征空间。这种严格的约束也有效地减少了本地数据邻域中固有的类不平衡,从而在本地雕刻出更平衡的类边界,并更有效地使用特征空间。在五个不同的数据集上进行了大量的实验,结果表明,我们的方法相比,在人脸识别和人的重新识别任务的最先进的模型的优越性。

英文摘要:

We present a novel framework to exploit privileged information for recognition which is provided only during the training phase. Here, we focus on recognition task where images are provided as the main view and soft biometric traits (attributes) are provided as the privileged data (only available during training phase). We demonstrate that more discriminative feature space can be learned by enforcing a deep network to adjust adaptive margins between classes utilizing attributes. This tight constraint also effectively reduces the class imbalance inherent in the local data neighborhood, thus carving more balanced class boundaries locally and using feature space more efficiently. Extensive experiments are performed on five different datasets and the results show the superiority of our method compared to the state-of-the-art models in both tasks of face recognition and person re-identification.

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。

↑