Neural Person Search Machines

Abstract

We investigate the problem of person search in the wild in this work. Instead of comparing the query against all candidate regions generated in a query-blind manner, we propose to recursively shrink the search area from the whole image till achieving precise localization of the target person, by fully exploiting information from the query and contextual cues in every recursive search step. We develop the Neural Person Search Machines (NPSM) to implement such recursive localization for person search. Benefiting from its neural search mechanism, NPSM is able to selectively shrink its focus from a loose region to a tighter one containing the target automatically. In this process, NPSM employs an internal primitive memory component to memorize the query representation which modulates the attention and augments its robustness to other distracting regions. Evaluations on two benchmark datasets, CUHK-SYSU Person Search dataset and PRW dataset, have demonstrated that our method can outperform current state-of-the-arts in both mAP and top-1 evaluation protocols.

Re-ID Driven Localization Refinement for Person Search

“…We compare our proposed network with current stateof-the-art methods including OIM [34], NPSM [22], RCAA [1], I-Net [13], MGTS [3], CLSA [17] on two popular datasets CUHK-SYSU [34] and PRW [41]. In addition to these methods, we also compare with the methods that joint different pedestrian detectors (DPM [8], ACF [6], CCF [35], LDCF [24], and R-CNN [10]) and person descriptors (BoW [39], LOMO [19], DenseSIFT-ColorHist (DSIFT) [38]) and distance metric (KISSME [16], XQDA [19]).…”

Learning Context Graph for Person Search

“…To facilitate real-world person re-id, recent methods propose to jointly address the task of detection and reidentification [43,42]. State-of-the-art methods [15,26,41] design online learning object functions to learn large number of identities in the training set. These methods achieve great performance on recent person search datasets.…”

Query-Guided End-To-End Person Search

“…More recently, [35] extends the OIM with an additional center loss [34], which improves the intra-class feature compactness. To our knowledge, the OIM loss is currently best for optimizing the joint network, adopted by most recent work [23,35], including ours. Query-guided person search.…”

“…Query-guided person search. To the best of our knowledge, the NPSM approach of Liu et al [23] is the sole to exploit the query image. They do so by instantiating an iterative person search mechanism based on a Conv-LSTM, which re-weights attention across a number of pre-defined person detection proposals.…”

“…To the best of our knowledge, across the person search approaches [23,37,39,42,46], only OIM [37] and IAN [35] optimize jointly the detector and the re-identification networks (end-to-end). On the other hand, NPSM [23] is the sole to adopt a query attention mechanism, by replacing the detector RPN with an iterative queryguided search based on Conv-LSTM. In fact NPSM builds on OIM but it is not end-to-end, since it freezes the detector and re-identification network parts to pre-trained values, arXiv:1905.01203v1 [cs.CV] 3 May 2019 i.e.…”