MOT Challenge 2015

Colab Notebook was used to perform the tests

As part of my professional residencies at the CIMAT, this research and testing was carried out to learn more about MOT algorithms and later use the information to use them in RGB-D images.

• Detections were obtained using Detectron2
• The algorithms for multiple object tracking was used are SORT and Deep SORT
• The MOT metrics were obtained by py-motmetrics

MOT Metrics with Detectron2 detections

• SORT (TUD-Campus Sequence)

              IDF1   IDP   IDR  Rcll  Prcn GT MT PT ML  FP FN IDs  FM  MOTA  MOTP IDt IDa IDm
  TUD-Campus 53.4% 48.3% 59.6% 83.6% 67.7%  8  5  3  0 143 59   8  10 41.5% 0.255   2   6   0
  OVERALL    53.4% 48.3% 59.6% 83.6% 67.7%  8  5  3  0 143 59   8  10 41.5% 0.255   2   6   0

• Deep SORT (TUD-Campus Sequence)

              IDF1   IDP   IDR  Rcll  Prcn GT MT PT ML FP FN IDs  FM  MOTA  MOTP IDt IDa IDm
  TUD-Campus 55.8% 56.4% 55.2% 76.6% 78.3%  8  3  5  0 76 84   8  10 53.2% 0.235   2   7   1
  OVERALL    55.8% 56.4% 55.2% 76.6% 78.3%  8  3  5  0 76 84   8  10 53.2% 0.235   2   7   1

Detectron2 video detections

SORT with Detectron2 detections - video

Deep SORT with Detectron2 detections - video