API documentation
Person Re-Identification API: integration guide
Create versioned 768-D person-appearance embeddings from individual crops, store them in a vector database, and design a human-reviewed retrieval workflow.
PolyReID field notes
Person ReID, explained for production teams
Clear engineering guidance for turning one cropped person into a versioned visual vector—and building the review, retrieval and privacy boundaries around it.
Knowledge base
Guides and field notes
Foundations
Person Re-Identification: a practical guide for production systems
Understand person ReID inputs, embeddings, evaluation, retrieval, failure modes and the human controls required beyond a model benchmark.
Foundations
Person ReID vs face recognition vs tracking
Compare person-appearance retrieval, facial recognition and video tracking by input, output, failure mode and governance impact.
Engineering
ReID embeddings and vector databases
Store, index and query normalized person-appearance embeddings without losing model compatibility, privacy boundaries or review context.
Case study
Ballroom Person ReID: what our internal evaluation does—and does not—show
A sourced account of PolyReID’s ballroom training setup, model design, workflow boundaries and known failure modes without universal accuracy claims.