Imperial College London

Professor Yiannis Demiris

Faculty of EngineeringDepartment of Electrical and Electronic Engineering

Professor of Human-Centred Robotics, Head of ISN



+44 (0)20 7594 6300y.demiris Website




1014Electrical EngineeringSouth Kensington Campus






BibTex format

author = {Choi, J and Chang, HJ and Fischer, T and Yun, S and Lee, K and Jeong, J and Demiris, Y and Choi, JY},
title = {Context-aware Deep Feature Compression for High-speed Visual Tracking},
url = {},
year = {2018}

RIS format (EndNote, RefMan)

AB - We propose a new context-aware correlation filter based tracking framework toachieve both high computational speed and state-of-the-art performance amongreal-time trackers. The major contribution to the high computational speed liesin the proposed deep feature compression that is achieved by a context-awarescheme utilizing multiple expert auto-encoders; a context in our frameworkrefers to the coarse category of the tracking target according to appearancepatterns. In the pre-training phase, one expert auto-encoder is trained percategory. In the tracking phase, the best expert auto-encoder is selected for agiven target, and only this auto-encoder is used. To achieve high trackingperformance with the compressed feature map, we introduce extrinsic denoisingprocesses and a new orthogonality loss term for pre-training and fine-tuning ofthe expert auto-encoders. We validate the proposed context-aware frameworkthrough a number of experiments, where our method achieves a comparableperformance to state-of-the-art trackers which cannot run in real-time, whilerunning at a significantly fast speed of over 100 fps.
AU - Choi,J
AU - Chang,HJ
AU - Fischer,T
AU - Yun,S
AU - Lee,K
AU - Jeong,J
AU - Demiris,Y
AU - Choi,JY
PY - 2018///
TI - Context-aware Deep Feature Compression for High-speed Visual Tracking
UR -
UR -
ER -