Benchmark:

MOT16 | MOT17 | MOT20 |

Short name:

FeatureSORT

Detector:

Public

Description:

We introduce a novel tracker designed for online multiple object tracking with a focus on using multiple feature modules each of which stands for a particular appearance information. By integrating distinct appearance features, including clothing color, style, and target direction, alongside a ReID network for robust
embedding extraction, our tracker achieves SOTA results.

Reference:

H. Hashempoor, R. Koikara, Y. Hwang. FeatureSORT: Essential Features for Effective Tracking. In , 2024.

Last submitted:

July 09, 2024 (5 months ago)

Published:

July 09, 2024 at 04:06:14 CET

Submissions:

1

Project page / code:

https://arxiv.org/abs/2407.04249v1

Open source:

Yes

Hardware:

2.1 GHZ, 8 Core

Runtime:

4.8 Hz

Benchmark performance:

Detailed performance:

Raw data:

download