Developers building machine learning or object detection models have to manually label hundreds to thousands of images to train them. To make the task easier, IBM has introduced a new auto-labeling tool.
The new feature is available for everyone through Cloud Annotations GUI, IBM’s online tool for labeling images. If you’re a developer, this is what you have to do:
- Head to the IBM Cloud Annotation tool from your desktop.
- Upload a few sample images from your dataset along with labels.
- Use instructions from here to train your model.
- Tap Autolabel in the interface and the tool will automatically annotate your images.
- You can review and edit labels.
Check out the video below to see the tool in action.
Cloud Annotations now has auto labeling 🎉 It's an early beta, but you can now upload a model to your project and let it do the labeling for you! pic.twitter.com/cEXJQVW2xl
— Nick Bourdakos (@bourdakos1) December 3, 2019
Nicholas Bourdakos, Developer Advocate, IBM Cognitive Applications, said the new auto labeling tool speeds up the process and allows developers to spend more time on other tasks:
Currently, it takes 200-500 samples of hand-labeled images for a model to detect one specific object. Autolableing images speeds the process and gives developers back valuable time to work on other innovative projects.
IBM has developed plenty of tools for you to get started with machine learning. You can pick a project from the Code Patterns page which features a tool from developing your own object recognition AI to a Generative Adversarial Network (GAN) to create your own fashion images.
You can read more about IBM’s auto-labeling tool here.