This repo provides object detection, object tracking and labeling functionality. The images used to develop the models contain a mix of public and private images. The private images are proprietary and not available for public consumption.
Acceptance criteria:
-Image slicer to cut images into nxn slices
-Zero padding of border slices
-Script and notebook to slice all images in a directory and save to output directory
For uploading AML labeled dataset into Custom Vision, enable updating an existing project with new tags. Currently, functionality only enables creating a new custom vision project.
Acceptance criteria:
New tags in AML labeled dataset can be uploaded to existing Custom Vision project
Blob-triggered Azure function that:
-calls custom vision to get bounding boxes
-crops image to extract region of interest
-send cropped image to computer vision API for dominant color