Ask Your Question

Revision history [back]

A large downside to an open source library with a BSD license is that once the techniques are integrated, companies can use it closed source and thus they do not need to make their models trained on their data publicly available. That being said, there is a set of models available in OpenCV, but it is rather limited.

  • For pedestrians the HOG+SVM model is embedded into OpenCV source code
  • For LatentSVM / DPM models, you can take a look here:

If you are willing to set up an online model sharing interface together with another researcher, I have been wondering to do so for the last 2 years (since my PhD research focusses on object detection models). Get in touch with me!