Exponential Triplet Loss

ICCDA 2020: Proceedings of the 2020 the 4th International Conference on Compute and Data Analysis 2020
Ēvalds Urtāns, Agris Ņikitenko, Valters Vēciņš

This paper introduces a novel variant of the Triplet Loss function that converges faster and gives better results. This function can separate class instances homogeneously through the whole embedding space. With Exponential Triplet Loss function we also introduce a novel type of embedding space regularization Unit-Range and Unit-Bounce that utilizes euclidean space more efficiently and resembles features of the cosine distance. We also examined factors for choosing the best embedding vector size for specific embedding spaces. Finally, we also demonstrate how new function can train models for one-shot learning and re-identification tasks.

Keywords
Feature embedding, Identification, One-shot learning, Re-identification, Sample mining, Triplet loss
DOI
10.1145/3388142.3388163
Hyperlink
https://dl.acm.org/doi/10.1145/3388142.3388163

Urtāns, Ē., Ņikitenko, A., Vēciņš, V. Exponential Triplet Loss. In: ICCDA 2020: Proceedings of the 2020 the 4th International Conference on Compute and Data Analysis, United States of America, San Jose, 9-12 March, 2020. New York: ACM, 2020, pp.152-158. ISBN 978-145037644-0. Available from: doi:10.1145/3388142.3388163

Publication language
English (en)