Vehicle Re-Identification: an Efficient Baseline Using Triplet Embedding

摘要: In this paper we tackle the problem of vehicle re-identification in a camera network utilizing triplet embeddings. Re-identification is matching appearances objects across different cameras. With proliferation surveillance cameras enabling smart and safer cities, there an ever-increasing need to re-identify vehicles Typical challenges arising city scenarios include variations viewpoints, illumination self occlusions. Most successful approaches for involve (deep) learning embedding space such that same identities are projected closer one another, compared representing identities. Popular loss functions (space) contrastive or loss. provide extensive evaluation these losses applied demonstrate using best practices embeddings outperform most previous proposed literature. Compared existing state-of-the-art approaches, our approach simpler more straightforward training only identity-level annotations, along with smallest published dimensions efficient inference. Furthermore work introduce formal sampling variant (batch sample) into