TY  - CPAPER
U1  - Konferenzveröffentlichung
A1  - Ho, Kalun
A1  - Pfreundt, Franz-Josef
A1  - Keuper, Margret
A1  - Keuper, Janis
T1  - Learning Embeddings for Image Clustering: An Empirical Study of Triplet Loss Approaches
T2  - Proceedings of ICPR 2020 : 25th International Conference on Pattern Recognition
N2  - 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.
KW  - correlation
KW  - pattern recognition
KW  - noise measurement
KW  - convolutional neural networks
KW  - image classification
Y1  - 2021
UR  - https://ieeexplore.ieee.org/abstract/document/9412602
UR  - https://www.computer.org/csdl/proceedings-article/icpr/2021/09412602/1tmjjjXo0xO
SN  - 1051-4651
SS  - 1051-4651
SN  - 978-1-7281-8808-9 (elektronisch)
SB  - 978-1-7281-8808-9 (elektronisch)
SN  - 978-1-7281-8809-6 (Print on Demand)
SB  - 978-1-7281-8809-6 (Print on Demand)
U6  - https://doi.org/10.1109/ICPR48806.2021.9412602
DO  - https://doi.org/10.1109/ICPR48806.2021.9412602
SP  - 87
EP  - 94
PB  - IEEE
ER  -