Particle Swarm Optimization

A Python implementation of Particle Swarm Optimization (PSO) — as suggested in [BrattonKennedy2007] — and its application to the Computer Vision task of Object Detection based on Template Matching.

Tested with:

matplotlib 3.3.2
numpy 1.19.2
python 3.8.5

Executable files

Run the files listed below to get the shown results. You will be asked whether to plot the animations or output textual information only.

• benchmarks.py

  PSO applied to non-convex artificial landscapes.

  

• detection.py

  PSO applied to Object Detection by Template Matching.

  

Reproducibility

PSO is a nondeterministic algorithm; however, for the sake of reproducibility of the results, the random seed in swarm.py is set to a fixed value — 42, for obvious reasons.

Conversely, if you want to get a different outcome at each run, replace

self.rng = np.random.default_rng(seed=42)

with

self.rng = np.random.default_rng()