/vision-tools

A bunch of cli tools for deep learning and computer vision.

Primary LanguagePythonMIT LicenseMIT

WARNING: IT IS NOT PROPERLY TESTED AND MAY DAMAGE YOUR FILES. USE AT YOUR OWN RISK.

vision-tools

A bunch of cli tools for deep learning and computer vision.

Modules

  • Extracts all faces in image with user defined padding from images.
  • Crops and saves all faces in image with correct orientation around Z axis.
  • Saves json file for each face with rotated and non-rotated rectangle, rotation angle.
  • Uses OpenCV DNN to run yolo model with *.weight, classes.names and *.cfg.
  • Takes list of class objects to detect and extracts the image, json file.

TODO

  • Replace python looping with numba njit parallel or use cython.
  • Make a setup script for install and run from terminal.
  • Upload object dataset generator code.
  • Use mediapipe face detector with face mesh landmarks as alternate of PCN.

References

Look into references.md.

License