In Defense of the Triplet Loss for Person Re-Identification
Indexed inarxivdatacite
Abstract
In the past few years, the field of computer vision has gone through a revolution fueled mainly by the advent of large datasets and the adoption of deep convolutional neural networks for end-to-end learning. The person re-identification subfield is no exception to this. Unfortunately, a prevailing belief in the community seems to be that the triplet loss is inferior to using surrogate losses (classification, verification) followed by a separate metric learning step. We show that, for models trained from scratch as well as pretrained ones, using a variant of the triplet loss to perform end-to-end deep metric learning outperforms most other published methods by a large margin.
Citation impact
2,895
total citations
- FWCI
- —
- Percentile
- —
- References
- 13
Citations per year
Authors
3Topics & keywords
Topics
Keywords
- Identification (biology)
- Psychology
- Computer science
- Biology
UN Sustainable Development Goals
- Peace, Justice and strong institutions
No related works found for this paper.