Automate Ground Truth Labeling for Object Detection
Version 220.127.116.11 (1.69 KB) by Alon Feldman
This file allows to use a pretrained Object Detection algorithm, to automate ground truth labeling in the Ground Truth Labeler app.
Updated 30 Nov 2021
To speed up the creation of a data base for training an Object Detector, we can use the automation feature in the MATLAB Image Labeler App. My goal was to automate the labeling process when training a custom YOLOv3 Object Detector.
Defining and using a custom automation algorithm requires you to create a class first. Once your class is created, save it into the appropriate folder. The instructions to do that are here.
Some algorithm types are available, such as vehicle detection and semantic segmentation, but there is no multi-label object detector.
This code had to be written, saved in a specific location, and run before it could be used.
With this code, you can open the Image Labler App, choose this automation algorithm, and use a pre-trained model to label more data faster.
Note that the Threshold for any algorithm is set as 0.5. you may change this number in the code if you like (on line 44).
Alon Feldman, Shai Kendler, Barak Fishbain (2021). Automate Ground Truth Labeling for Object Detection (https://www.mathworks.com/matlabcentral/fileexchange/102689), MATLAB Central File Exchange. Retrieved November 30, 2021.
MATLAB Release Compatibility
Created with R2021b
Compatible with R2021b and later releases
Platform CompatibilityWindows macOS Linux
Inspired by: Computer Vision Toolbox Model for YOLO v3 Object Detection
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!Start Hunting!
Discover Live Editor
Create scripts with code, output, and formatted text in a single executable document.