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 2007.03123cs.CV
收藏

图像聚类中的学习嵌入:三重损失方法的实证研究

Learning Embeddings for Image Clustering: An Empirical Study of Triplet Loss Approaches

Fraunhofer ITWM(弗劳恩霍夫工业数学研究所);University of Mannheim(曼海姆大学);Offenburg University(奥芬堡大学)

Kalun Ho, Janis Keuper, Franz-Josef Pfreundt, Margret Keuper

摘要:

在这项工作中,我们评估两个不同的图像聚类目标,k-均值聚类和相关聚类,在三重损失引起的特征空间嵌入的背景下。具体来说,我们训练一个卷积神经网络,通过优化两个流行版本的三重损失来学习判别特征,以便在噪声标签的假设下研究它们的聚类特性。此外,我们提出了一个新的,简单的三重损失制定,它显示出理想的性能方面的正式聚类目标,并优于现有的方法。我们评估所有三个三重损失公式的K均值和相关聚类CIFAR-10图像分类数据集。

英文摘要:

In this work, we evaluate two different image clustering objectives, k-means clustering and correlation clustering, in the context of Triplet Loss induced feature space embeddings. Specifically, we train a convolutional neural network to learn discriminative features by optimizing two popular versions of the Triplet Loss in order to study their clustering properties under the assumption of noisy labels. Additionally, we propose a new, simple Triplet Loss formulation, which shows desirable properties with respect to formal clustering objectives and outperforms the existing methods. We evaluate all three Triplet loss formulations for K-means and correlation clustering on the CIFAR-10 image classification dataset.

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。

↑