Skip to content

Pre-trained custom HOG detector for Cozmo cubes on greater distance

License

Notifications You must be signed in to change notification settings

maxosprojects/cozmo-far-cube-detector

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 

Repository files navigation

A pre-trained custom HOG (Histogram of Oriented Gradients) detector to detect Cozmo cubes on a greater distance than SDK is capable of.

The detector needs dlib installed, e.g.:

pip3 install --user dlib

Visualization of the pre-trained model (detector):

Visualization

The example.py displays Cozmo's video feed with bounding boxes for cubes. It does not detect which cube it is, only bounding boxes (if you look at Cozmo's feed the markers on the cubes indeed are too noisy at distance to be identified).

There is a faster HOG detector implementation that I'll be looking to implement in python, but even dlib is good enough with its ~200ms/image on a reasonably performant computer. Another improvement could be done by employing cubes' lights and some simple opencv functions.

About

Pre-trained custom HOG detector for Cozmo cubes on greater distance

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages