vltanh / trackun

A Python package for (multiple) object tracking using recursive Bayesian filtering

trackun

A Python package for (multiple) object tracking using recursive Bayesian filtering

Demo

Note: You can find more in images.

Example 1: GM-Bernoulli filter for single-object tracking on Constant Velocity with Gaussian noise model

python demo.py -s -m linear_gaussian -f GM-Bernoulli -o vis/output

Click to expand

Example 2: GM-PHD and GM-CPHD filter for multi-object tracking on Constant Velocity with Gaussian noise model

python demo.py -m linear_gaussian -f GM-PHD GM-CPHD -o vis/output

Click to expand

Example 3: GM-LMB filter for multi-object tracking on Constant Velocity with Gaussian noise model

python demo.py -m linear_gaussian -f GM-LMB -o vis/output

Click to expand

Checklist

Click to expand

Filters

Click to expand

• Single Object
  • Kalman Filter (GMS)
  • Particle Filter (SMC)
  • Extended Kalman Filter (EKF)
  • Unscented Kalman Filter (UKF)
• Bernoulli
  • Kalman Filter (GMS)
  • Particle Filter (SMC)
  • Extended Kalman Filter (EKF)
  • Unscented Kalman Filter (UKF)
• Probability Hypothesis Density (PHD)
  • Kalman Filter (GMS)
  • Particle Filter (SMC)
  • Extended Kalman Filter (EKF)
  • Unscented Kalman Filter (UKF)
• Cardinalized Probability Hypothesis Density (CPHD)
  • Kalman Filter (GMS)
  • Particle Filter (SMC)
  • Extended Kalman Filter (EKF)
  • Unscented Kalman Filter (UKF)
• Robust Probability Hypothesis Density (PHD)
  • Unknown clutter (Lambda-CPHD)
    • Kalman Filter (GMS)
    • Particle Filter (SMC)
    • Extended Kalman Filter (EKF)
    • Unscented Kalman Filter (UKF)
  • Unknown detection probability (pD-CPHD)
    • Kalman Filter (GMS)
    • Particle Filter (SMC)
    • Extended Kalman Filter (EKF)
    • Unscented Kalman Filter (UKF)
  • Unknown clutter rate and detection probability
    • Kalman Filter (GMS)
    • Particle Filter (SMC)
    • Extended Kalman Filter (EKF)
    • Unscented Kalman Filter (UKF)
• Cardinality Balanced Multi-target Multi-Bernoulli (CBMeMBer)
  • Kalman Filter (GMS)
  • Particle Filter (SMC)
  • Extended Kalman Filter (EKF)
  • Unscented Kalman Filter (UKF)
• Generalized Labeled Multi-Bernoulli (GLMB)
  • Kalman Filter (GMS)
  • Particle Filter (SMC)
  • Extended Kalman Filter (EKF)
  • Unscented Kalman Filter (UKF)
• Labeled Multi-Bernoulli (LMB)
  • Kalman Filter (GMS)
  • Particle Filter (SMC)
  • Extended Kalman Filter (EKF)
  • Unscented Kalman Filter (UKF)

Models

Click to expand

Motion model

• Linear Gaussian
  • Constant velocity
• Non-Linear Gaussian
  • Coordinated turn (CT)

Measurement model

• Linear Gaussian
  • Cartesian coordinate
• Non-Linear
  • Bearing/Polar coordinate

Other models

• Birth model
  • Multi-Bernoulli Gaussian
  • Multi-Bernoulli Gaussian Mixture
• Detection model
  • Constant probability
  • Bearing Gaussian
• Survival model
  • Constant probability
• Clutter model
  • Uniform clutter

Metrics

• OSPA
• OSPA2

Utility

• Examples and Visualization
• Benchmarking
• Optimization (consider memory-speed tradeoffs, JIT,...)
• System design and folder structure
• Testing

Credits

Original MATLAB implementation comes from http://ba-tuong.vo-au.com/codes.html