Smart Multiple Affinity Tracking [on] [SMAT]

Submitted on 12 Oct. 2019 00:23 by
Nicolas Franco Gonzalez (Grenoble INP)

Running time:0.1 s
Environment:1 core @ 2.5 Ghz (C/C++)

Method Description:
Tracking using optical flow, appearance vector,
intersection over union and pose estimation.
Hungarian method association
Latex Bibtex:
author="Gonzalez, Nicolas Franco
and Ospina, Andres
and Calvez, Philippe",
editor="Campilho, Aur{\'e}lio
and Karray, Fakhri
and Wang, Zhou",
title="SMAT: Smart Multiple Affinity Metrics for
Multiple Object Tracking",
booktitle="Image Analysis and Recognition",
publisher="Springer International Publishing",
abstract="This research introduces a novel
multiple object tracking algorithm called SMAT
(Smart Multiple Affinity Metric Tracking) that
works as an online tracking-by-detection
approach. The use of various characteristics from
observation is established as a critical factor
for improving tracking performance. By using the
position, motion, appearance, and a correction
component, our approach achieves an accuracy
comparable to state of the art trackers. We use
the optical flow to track the motion of the
objects, we show that tracking accuracy can be
improved by using a neural network to select key
points to be tracked by the optical flow. The
proposed algorithm is evaluated by using the
KITTI Tracking Benchmark for the class CAR.",

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 71.88 % 72.13 % 72.13 % 74.43 % 87.33 % 74.77 % 88.30 % 87.19 %

Benchmark TP FP FN
CAR 29138 5254 175

CAR 83.64 % 85.89 % 84.21 % 198 71.68 %

Benchmark MT rate PT rate ML rate FRAG
CAR 62.77 % 31.23 % 6.00 % 294

Benchmark # Dets # Tracks
CAR 29313 751

This table as LaTeX