synthetic image generator for object detection/segmentation network training
This tool was developed for utilizing the opencv interface to generate synthetic image(with 3d transform) samples with alternative background for object detection/segmentation network training. Currently, it contains alternative implementions including c++ and python. c++ version will released later.
For previous project, i write this script to generate huge size of synthetic ocr images with random background as well as the corresponding annotation files. It works fine and helps a lot to the network training task.
Hope it will be helpful.
Contact Author: tower.zhang [email protected]
License synthetic-image-generator is licensed is licensed under GPL 3.0.
Software Requirements:
python3 numpy csv cv2