Abstract
Retrieving the given objects hidden amidst the gallery set is important for public safety and decision-making. Heterogeneous pedestrian retrieval (person re-identification) aims to retrieve the same person images from different modality set for identification. To address this problem, we contribute a new character-illustration-style image and normal photo pedestrian re-identification dataset (CINPID), which is collected on campus. The CINPID dataset includes two modalities, i.e., normal photos captured by one camera and character-illustration-style images drawn by the painter. To handle the problem of pedestrian retrieval with character-illustration-style images and normal photos, we propose a semi-coupled mapping and discriminant dictionary learning (SMD <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> L), which can learn a semi-coupled mapping matrix and dictionary pair from heterogeneous samples. With the learned semi-coupled mapping matrix, the differences between heterogeneous data can be reduced to some extent. Experimental results on the new CINPID dataset show that our approach outperforms the compared competing methods.
Highlights
Identifying the given objects hidden amidst the gallery set plays an important role in recognition [2] and re-identification [4], [5]
To the best of our knowledge, it is the first illustration-style pedestrian dataset for person retrieval problem. iii) We propose a semi-coupled mapping and discriminant dictionary learning (SMD2L) for CTNPR problem, which can learn a semi-coupled matrix and dictionary pair
The main advantages of our approach have four-fold: i) SMD2L employs the semi-coupled mapping strategy, which can bridge the relationship of the illustration-style images and normal photos, and reduce the differences between heterogeneous samples. ii) The learned metric matrix can uncover the intrinsic projection of heterogeneous data. iii) The dictionary pair can be favorable for illustration-style images and normal photos under complex scenario respectively. iv) The discriminant constraint can make the same class compact and different classes separated away, which is favorable for retrieval
Summary
Identifying the given objects hidden amidst the gallery set plays an important role in recognition [2] and re-identification [4], [5]. Ii) There exist significant differences in the viewpoints between illustration-style image and normal photo, which will be difficult for pedestrian retrieval. SYSU-MM01 [5] is introduced by RGB camera and near-infrared (NIR) camera for NIR person re-id, which is suitable under low illumination scene There exist several dictionary learning based heterogeneous pedestrian retrieval methods. Cross-view quadratic discriminant analysis (XQDA) [36] is presented to solve person re-id task under normal scene, which can learn a discriminant low dimensional subspace by cross-view quadratic discriminant analysis and a metric on the derived subspace They design a method Local Maximal Occurrence (LOMO) to represent well person, which can make a stable representation against viewpoint changes. To the best of our knowledge, there is no illustration-style image-based pedestrian retrieval dataset to date
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
