HGFMOT: optimize the backbone structures and loss functions based on anchor free framework

MOT16-01


Benchmark:

MOT17 | MOT16 |

Short name:

HGFMOT

Detector:

Private

Description:

Reference:

submission by AI Lab, Lenovo Research

Last submitted:

January 12, 2021 (10 days ago)

Published:

November 26, 2020 at 11:30:10 CET

Submissions:

3

Project page / code:

n/a

Open source:

No

Hardware:

V100

Runtime:

9.9 Hz

Benchmark performance:

Sequence MOTA IDF1 MOTP MT ML FP FN Recall Precision FAF ID Sw. Frag
MOT1677.674.981.0362 (47.7)93 (12.3)9,62030,07383.594.11.61,067 (12.8)2,201 (26.4)

Detailed performance:

Sequence MOTA IDF1 MOTP MT ML FP FN Recall Precision FAF ID Sw. Frag
MOT16-0162.863.480.91114981,82471.590.21.154145
MOT16-0391.083.981.713413,3395,88394.496.72.2157499
MOT16-0663.665.979.897437893,19972.391.40.7207244
MOT16-0768.065.780.42519394,17874.492.81.9107302
MOT16-0852.745.880.92122,0865,57666.784.33.3248422
MOT16-1256.264.182.429149602,64068.285.51.135150
MOT16-1456.567.176.445311,0096,77363.492.11.3259439

Raw data: