Python program to convert OpenImages (V4/V5) labels to be used for YOLOv3. The program is a more efficient version (15x faster) than the repository by Karol Majek..
The program can be used to train either for all the 600 classes or for few classes (for custom object detection models).
-
Download the annotation .csv files as well as the class description .csv file.
-
Download the images from the OpenImages dataset. For one or few classes, check this repository.