A semi-supervised person re-identification method that progressively generates cross-camera soft labels from an affinity matrix and trains with weighted cross-entropy or triplet losses reaches near-supervised accuracy without cross-camera labels.
Learning deep feature representations with domain guided dropout for person re-identification,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Progressive Cross-camera Soft-label Learning for Semi-supervised Person Re-identification
A semi-supervised person re-identification method that progressively generates cross-camera soft labels from an affinity matrix and trains with weighted cross-entropy or triplet losses reaches near-supervised accuracy without cross-camera labels.