Title Video-based person re-identification


Person Re-ID in multi-camera surveillance system has been researched increasingly in recent years due to its presence in numerous important applications, such as surveillance, forensics, smart environments, multimedia applications, and so on [1]. It is a problem of person identity association across camera views at different locations and times. Most of the reported person

Re-ID methods deal with one sole image and the manually extracted human ROIs (Region of Interest) [2]. Therefore, rich information containing in image sequence (video) is still under-exploited. Moreover, when working with automatic ROI detection, we have to face to two open issues of object detection: miss detection and false positive. Recently, deep learning has been applied for video-based person re-identification [3]. The main aim of this work is to investigate deeply the use of deep learning for video-based person re-identification.


Work description:

-   Study person re-identification

-   Study different deep learning architecture for person re-identification.

-   Propose suitable framework for person re-identification based on deep learning

-   Evaluate the proposed method in available datasets such as MARS dataset


Student prerequisites

This subject is dedicated to Vietnamese students as well as foreigner. The students who have a fairly good knowledge about computer vision, image processing and deep learning are privileged.


Ass.Prof. LE Thi Lan: This email address is being protected from spambots. You need JavaScript enabled to view it.


[1] A. Bedagkar-Gala, S. K. Shah, A survey of approaches and trends in person  re-identification, Image and Vision Computing 32 (4) (2014) 270-286.

[2] Thanh-Thuy Pham, Thi-Lan Le, Trung-Kien Dao, Duy-Hung Le, A robust model for person re-identification in multimodal person localization, in The Ninth International Conference on Mobile Ubiquitous Computing, Systems, Services and Technologies (UBICOMM 2015), 2015, p. 51

[3] Lin Wu, Chunhua Shen, Anton van den Hengel,  Convolutional LSTM Networks for Video-based Person Re-identification, 2016.