English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The goal of person search is to localize and match query persons from sceneimages. For high efficiency, one-step methods have been developed to jointlyhandle the pedestrian detection and identification sub-tasks using a singlenetwork. There are two major challenges in the current one-step approaches. Oneis the mutual interference between the optimization objectives of multiplesub-tasks. The other is the sub-optimal identification feature learning causedby small batch size when end-to-end training. To overcome these problems, wepropose a decoupled and memory-reinforced network (DMRNet). Specifically, toreconcile the conflicts of multiple objectives, we simplify the standardtightly coupled pipelines and establish a deeply decoupled multi-task learningframework. Further, we build a memory-reinforced mechanism to boost theidentification feature learning. By queuing the identification features ofrecently accessed instances into a memory bank, the mechanism augments thesimilarity pair construction for pairwise metric learning. For better encodingconsistency of the stored features, a slow-moving average of the network isapplied for extracting these features. In this way, the dual networks reinforceeach other and converge to robust solution states. Experimentally, the proposedmethod obtains 93.2% and 46.9% mAP on CUHK-SYSU and PRW datasets, which exceedsall the existing one-step methods.

Decoupled and Memory-Reinforced Networks: Towards Effective Feature Learning for One-Step Person Search