Pre-trained priors from foundation models drive substantial performance variance in person re-identification, with simple fine-tuning of large models reaching SOTA results while staying close to initial parameters.
The open images dataset v4.International Journal of Com- puter Vision, 128(7):1956–1981
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Not All Starting Points Are Equal: Pre-trained Priors and Their Outsized Impact on Person Identification
Pre-trained priors from foundation models drive substantial performance variance in person re-identification, with simple fine-tuning of large models reaching SOTA results while staying close to initial parameters.