Method

TDFL with Recurrent Rolling Convolution [RRC] car detections [on] [TDFL*]


Submitted on 4 Jan. 2019 02:38 by
Handuo Zhang (Nanyang Technological University)

Running time:0.01 s
Environment:1 core @ 2.5 Ghz (C/C++)

Method Description:
This work has been accepted by 2019 IEEE 15th
International Conference on Control and
Automation. Our method makes use of a
dissimilarity measure
calculated based on object location measure and
object appearance measure. These dissimilarity
values are used in Hungarian Algorithm in the
data association step for track identity
assignment. Location measure is calculated using
the predicted object location and bounding box,
while the appearance measure is from the last
feature layer from the detection network. Main
focus in this work is to propose a tracking
framework that can be used in real time automated
vehicle guiding applications, by striking a
balance between computational complexity and
tracking accuracy. Therefore, we make use of the
deep features available from detection framework
rather than calculating a new appearance measure
during the tracking step. The method propose is
very efficient and enables to achieve speeds up
to 500+ frames per second (fps) in KITTI tracking
benchmark while achieving state-of-the art
results.
Parameters:
\alpha\_f=0.5, \alpha\_l=0.5
Latex Bibtex:

Detailed Results

From all 29 test sequences, our benchmark computes the commonly used tracking metrics CLEARMOT, MT/PT/ML, identity switches, and fragmentations [1,2]. The tables below show all of these metrics.


Benchmark MOTA MOTP MODA MODP
CAR 84.30 % 85.63 % 85.26 % 88.78 %

Benchmark recall precision F1 TP FP FN FAR #objects #trajectories
CAR 88.46 % 97.95 % 92.97 % 33517 700 4371 6.29 % 38245 830

Benchmark MT PT ML IDS FRAG
CAR 71.69 % 25.38 % 2.92 % 328 815

This table as LaTeX


[1] K. Bernardin, R. Stiefelhagen: Evaluating Multiple Object Tracking Performance: The CLEAR MOT Metrics. JIVP 2008.
[2] Y. Li, C. Huang, R. Nevatia: Learning to associate: HybridBoosted multi-target tracker for crowded scene. CVPR 2009.


eXTReMe Tracker