ICPR26 TRVID competition on Privacy-Preserving Person Re-Identification Call for Papers
ICPR26 TRVID competition on Privacy-Preserving Person Re-Identification
We invite you to participate in TVRID, the ICPR 2026 Competition on
Privacy-Preserving Person Re-Identification from Top-View RGB-Depth Camera.
Overview (dataset and setting).
TVRID is a top-view person re-identification benchmark with aligned RGB and
Depth sequences, captured using four overhead Intel RealSense D455 cameras.
The dataset includes 88 identities, with each person observed twice (IN/OUT),
across four geometric contexts: flat ground, ascent, descent, and an oblique
roof view. This setup is designed to support research on privacy-preserving
Re-ID (Depth) and cross-modal RGB?Depth retrieval in a realistic
multi-camera overhead scenario.
Who is it for? Researchers and practitioners working on:
* Person Re-ID / metric learning / retrieval
* Cross-modal learning
* Privacy-preserving computer vision
* Sequence modeling for Re-ID
Tracks.
* RGB Re-ID (privacy-unconstrained)
* Depth Re-ID (privacy-preserving)
* Cross-modal RGB?Depth retrieval
Evaluation.
Submissions are ranked lists evaluated with CMC@1/5/10 and mAP. The primary
leaderboard metric is overallMap (mean of per-track overall mAP).
The top five teams will be invited to contribute to the competition summary
paper to be published in the ICPR 2026 proceedings.
How to participate.
* Competition page (registration, submission, rules):
https://www.codabench.org/competitions/12200/#/pages-tab
* Baseline code and instructions (TVRID quick start, data preparation,
and submission format): https://github.com/RaphaelDel/ICPR-2026-TVRID
If you have colleagues or students working on person Re-ID, privacy-preserving
vision, metric learning, or cross-modal retrieval, please feel free to
forward this email.
Contact: Raphael Delecluse