Method

CAMO-MOT[la] [CAMO-MOT]


Submitted on 1 Nov. 2021 14:48 by
Wenyuan Qin (Tsinghua University)

Running time:0.04 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
In this paper we propose a novel multi-modal
fusion framework called CAMO. Our method uses
Optimal Occlusion State-based Object Apperance
Module to achieve the selection of optimal object
appearance features, Confidence Score-based Motion
Module to achieve the localization and prediction
of objects in 3D space, and Multi Category Multi-
Modal Fusion Association Module achieves stable
association of multiple categories under multi-
modal conditions.
Parameters:
\thea_post=8.5 \theta_aapperance=0.01
\theta_motion=10.3
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 79.99 % 76.34 % 84.45 % 81.16 % 84.59 % 87.27 % 90.30 % 86.66 %
PEDESTRIAN 44.77 % 41.53 % 48.70 % 45.16 % 60.25 % 55.35 % 59.63 % 71.22 %

Benchmark TP FP FN
CAR 32055 2337 942
PEDESTRIAN 14825 8325 2525

Benchmark MOTA MOTP MODA IDSW sMOTA
CAR 90.38 % 85.00 % 90.47 % 30 76.40 %
PEDESTRIAN 52.48 % 64.50 % 53.13 % 152 29.74 %

Benchmark MT rate PT rate ML rate FRAG
CAR 84.46 % 8.00 % 7.54 % 156
PEDESTRIAN 35.40 % 38.83 % 25.77 % 1133

Benchmark # Dets # Tracks
CAR 32997 686
PEDESTRIAN 17350 311

This table as LaTeX