Skip to content

utilities for working on image/label files for machine learning training

Notifications You must be signed in to change notification settings

jlwebuser/imageutil

Repository files navigation

Utilties for working on image/label data for ML

For detailed usage just specify --help on the command line for any file.

cleanupnames.py - scan a directory for all .jpg and associated yolo_mark .txt files and copy to a new directory with a new base filename +integer.

detectimg.py - Run detection on an MxNet network drawing bounding boxes of classes detected in model in the images specified. Numerous options to control threshold, images, display/recording of detection.

mirrorxyz.py - make a mirror copy of a jpg around the x, y, or z axis (to get a reflection to increase) and if there is a yolo_mark label file of the same name, created a mirror of it also.

makelst.py - reads a directoy of jpg and txt files that are of the yolo_mark format transforms to stdout a LST file that can be input to im2rec.py so you can make a mxnet REC file for training or validation...

resizeimg.py - Resize all jpg files in a directory preserving aspect ratio.

dvr.py - Class implements a simple recorder for short clips into an mp4 file. It is built to record a short clip of recgonition images after a network has detected an object. Once triggered it runs for a specified duration to get context after a detection occurrs.

About

utilities for working on image/label files for machine learning training

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages