In Defense of the Triplet Loss for Person Re-Identification
@article{Hermans2017InDO, title={In Defense of the Triplet Loss for Person Re-Identification}, author={Alexander Hermans and Lucas Beyer and Bastian Leibe}, journal={ArXiv}, year={2017}, volume={abs/1703.07737} }
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… CONTINUE READING
Figures, Tables, and Topics from this paper.
Citations
Publications citing this paper.
SHOWING 1-10 OF 445 CITATIONS, ESTIMATED 99% COVERAGE
Re-identification of Vessels with Convolutional Neural Networks
VIEW 7 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED
Single Camera Training for Person Re-identification
VIEW 7 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED
Deep Cosine Metric Learning for Person Re-identification
VIEW 10 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED
Homocentric Hypersphere Feature Embedding for Person Re-Identification
VIEW 6 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED
Video Person Re-Identification by Temporal Residual Learning
VIEW 4 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED
Adversarial Camera Alignment Network for Unsupervised Cross-camera Person Re-identification
VIEW 4 EXCERPTS
CITES BACKGROUND, METHODS & RESULTS
HIGHLY INFLUENCED
CARF-Net: CNN attention and RNN fusion network for video-based person reidentification
VIEW 6 EXCERPTS
CITES METHODS
HIGHLY INFLUENCED
CamStyle: A Novel Data Augmentation Method for Person Re-Identification
VIEW 5 EXCERPTS
CITES METHODS
HIGHLY INFLUENCED
Full batch loss for person re-identification
VIEW 10 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED
Hdrnet: Person Re-Identification Using Hybrid Sampling in Deep Reconstruction Network
VIEW 6 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED
FILTER CITATIONS BY YEAR
CITATION STATISTICS
86 Highly Influenced Citations
Averaged 147 Citations per year from 2017 through 2019
28% Increase in citations per year in 2019 over 2018
References
Publications referenced by this paper.
SHOWING 1-10 OF 49 REFERENCES
Scalable Person Re-identification: A Benchmark
VIEW 17 EXCERPTS
HIGHLY INFLUENTIAL
Deep Sparse Rectifier Neural Networks
VIEW 4 EXCERPTS
HIGHLY INFLUENTIAL
Deep Metric Learning via Lifted Structured Feature Embedding
VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL
Deep Transfer Learning for Person Re-Identification
VIEW 8 EXCERPTS
HIGHLY INFLUENTIAL
Distance Metric Learning for Large Margin Nearest Neighbor Classification
VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL
Embedding Deep Metric for Person Re-identification: A Study Against Large Variations
VIEW 7 EXCERPTS
HIGHLY INFLUENTIAL
Joint Learning of Single-Image and Cross-Image Representations for Person Re-identification
VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL
MARS: A Video Benchmark for Large-Scale Person Re-Identification
VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL
Person Re-identification by Multi-Channel Parts-Based CNN with Improved Triplet Loss Function
VIEW 6 EXCERPTS
HIGHLY INFLUENTIAL
DeepReID: Deep Filter Pairing Neural Network for Person Re-identification
VIEW 8 EXCERPTS
HIGHLY INFLUENTIAL