Method

Observation-Centric SORT [OC-SORT]
[Anonymous Submission]

Submitted on 6 Mar. 2022 04:09 by
[Anonymous Submission]

Running time:0.01 s
Environment:8 cores @ 3.0 Ghz (Python)

Method Description:
An online method with pure motion model. No appearance features or
other cues are used. No heuristic post-processing is required. Given
detections, the inference speed is 793FPS on a Intel i9-9980XE CPU
@ 3.00GHz.
Parameters:
We directly uses the detection results from permaTrack with the
model trained on partial training data (available on its github repo:
https://github.com/TRI-ML/permatrack).
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 70.22 % 71.97 % 69.44 % 77.20 % 81.00 % 74.31 % 84.17 % 84.40 %
PEDESTRIAN 53.22 % 48.72 % 58.39 % 53.56 % 70.83 % 63.67 % 72.82 % 78.91 %

Benchmark TP FP FN
CAR 31495 2897 1284
PEDESTRIAN 15933 7217 1573

Benchmark MOTA MOTP MODA IDSW sMOTA
CAR 87.01 % 82.28 % 87.84 % 286 70.78 %
PEDESTRIAN 61.12 % 75.18 % 62.03 % 211 44.04 %

Benchmark MT rate PT rate ML rate FRAG
CAR 77.69 % 18.92 % 3.38 % 173
PEDESTRIAN 40.21 % 33.68 % 26.12 % 423

Benchmark # Dets # Tracks
CAR 32779 756
PEDESTRIAN 17506 294

This table as LaTeX