AFJPDA: Appearance Feature-aided JPDA

1. Software Introduction

This is a multi-class multi-object tracking (MCMOT) algorithm with joint probabilistic data association (JPDA) filter based on FairMOT and MCMOT on following links:

FairMOT MCMOT

Detail of the algorithm can be found from following paper: AFJPDA

Developed by Sukkeun Kim

• Email: s.kim.aero@gmail.com

2. Running the Demo

1. Setup the environment following the FairMOT repository.
2. Run by following commands:

   $ conda activate AFJPDA $ cd src $ python demo.py --load_model ../Your_pretrained_model.pth --input-video ../Your_test_video.mp4 --id_weight 2 --conf_thres 0.4

• Examples:

  $ python demo.py --load_model ../exp/models/mcmot_last_track_dla_34_carla_64000.pth --input-video ../exp/videos/Test.mp4 --id_weight 2 --conf_thres 0.4 $ python demo.py --load_model ../exp/models/mcmot_last_track_dla_18_visdrone.pth --input-video ../exp/videos/Test_visdrone.mp4 --id_weight 2

• Note: id_weight 0 for detection only, 1 for MCMOT by Even, 2 for JPDA, and 3 for AFJPDA

3. Using Own Dataset

• Need to check below two files for using other dataset:

  • opts.py file in src/lib: Number of classes is defined here
  • gen_dataset_yourdataset: Class IDs are defined here (Check multitracker.py file)

• Training

  $ python train.py

• Training data label: [ClassID, ReID, X, Y, W, H]

4. Version Information

• 1st Mar 2023 Beta 0.0: First commit
• 30th Mar 2023 Beta 1.0: First full

--- Older versions are not available on github ---

• 19th Jul 2024 Release 1.0: First public