DART 3 : Leveraging Distance for Test Time Adaptation in Person Re-Identification
Rajarshi Bhattacharya, Shakeeb Murtaza, Christian Desrosiers, Jose Dolz, Maguelonne Heritier, Eric Granger · 2025
Read DART 3 : Leveraging Distance for Test Time Adaptation in Person Re-Identification by Rajarshi Bhattacharya, Shakeeb Murtaza, Christian Desrosiers, Jose Dolz, Maguelonne Heritier, Eric Granger (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-18337.txt.
Description
Rajarshi Bhattacharya and five colleagues help security cameras recognize the same person across new cameras. Person re-identification models often learn camera quirks instead of people, so adding a camera wrecks accuracy. Fixing that at test time, without the old training data, is the goal.
Their method tunes itself using distances between images, not class guesses, since retrieval is a distance game. A match's nearest neighbor gap predicts its error, which steers the update. No retraining, source data, or model changes are needed, and a light version trims the cost. Across standard benchmarks, the approach beat other test-time methods. The authors offer it for live surveillance upgrades.
Catalog record
| CREATOR | Rajarshi Bhattacharya, Shakeeb Murtaza, Christian Desrosiers, Jose Dolz, Maguelonne Heritier, Eric Granger |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-18337