Method

Bilateral Association Tracking [on] [BAT]


Submitted on 5 Mar. 2022 10:41 by
You Yuan Chen (Beijing University Of Chemical Technology)

Running time:0.03s
Environment:6core @ 4.5 Ghz (Python)

Method Description:
A Bilateral Association Tracking(BAT) framework .
It uses tracklets as the basic node instead of
discrete detection for tracking. Meanwhile, a
Parzen density based Hierarchical Agglomerative
Clustering algorithm is introduced to describe the
density distribution of targets and generate
tracklets with high confidence.
Parameters:
\alpha=1.5,2.0,2.5
\beta=1.5,2.0,2.5
Latex Bibtex:

Detailed Results

From all 29 test sequences, our benchmark computes the HOTA tracking metrics (HOTA, DetA, AssA, DetRe, DetPr, AssRe, AssPr, LocA) [1] as well as the CLEARMOT, MT/PT/ML, identity switches, and fragmentation [2,3] metrics. The tables below show all of these metrics.


Benchmark HOTA DetA AssA DetRe DetPr AssRe AssPr LocA
CAR 68.49 % 71.53 % 66.14 % 75.24 % 83.89 % 70.51 % 83.69 % 85.26 %
PEDESTRIAN 42.91 % 43.91 % 42.22 % 47.19 % 71.11 % 45.25 % 73.40 % 77.92 %

Benchmark TP FP FN
CAR 30358 4034 485
PEDESTRIAN 14298 8852 1064

Benchmark MOTA MOTP MODA IDSW sMOTA
CAR 86.20 % 83.28 % 86.86 % 227 71.44 %
PEDESTRIAN 55.71 % 73.71 % 57.17 % 338 39.47 %

Benchmark MT rate PT rate ML rate FRAG
CAR 72.61 % 19.08 % 8.31 % 304
PEDESTRIAN 30.93 % 38.14 % 30.93 % 589

Benchmark # Dets # Tracks
CAR 30843 688
PEDESTRIAN 15362 368

This table as LaTeX