Looking for help getting started with HyperLabel, our desktop application for creating labeled datasets for machine learning (ML) quickly, easily and with the ability to keep your data private? Well, your search is over.
HyperLabel has just launched a series of practical, step-by-step tutorials intended to help users quickly become HyperLabel “power users”. The full text of the HyperLabel Fundamentals series – complete with step-by-step advice and screenshot visuals – is being published on the new Sixgill Medium channel. The first post in the series – “HyperLabel Fundamentals: Image Classification with Create ML” – is available now.
Be sure to follow the Sixgill Medium Channel for more on HyperLabel fundamentals as well as the latest trends and insights on vision AI, continuous training, MLOps and related topics. And remember, the HyperLabel Developer version is available free, with no label quantity restrictions! If you haven’t already, you can download it from the App Store, or Microsoft.
The first HyperLabel fundamentals post offers step-by-step help navigating the process of assigning classifications to images within HyperLabel. The tutorial is built around an example that walks readers through setting up a project for training an ML model to recognize and classify different shark species from a collection of uploaded images.
Here’s a brief overview of topics covered in the tutorial:
Gathering Your Images
The tutorial starts with advice on how to obtain images for your classification project. A simple (and free) way for developers to gather photos for our sample shark classification project, for example, is to use the Bing Image Search API, which Microsoft makes available to developers free for limited uses. There’s no need to specify license criteria in the API call, since using copyrighted material to train machine learning models is permissible.
One important step, however, is to select Photos as the image type. Failing to do so would result in images being retrieved that are illustrations not appropriate for ML labeling.
The tutorial provides all of the information you need to use HyperLabel for labeling images so they are ready for use to train ML models. You can, for example, use HyperLabel to filter out irrelevant images by using the filtering settings.
Setting Up Your Image Classification Project
As the tutorial notes, creating a new project from HyperLabel’s Create Project screen is quick and easy. Simply hit Create Project and you’re on your way. The fundamentals tutorial will take you through the process of creating a project, as well as an image directory, and creating a customized labeling schema for specific projects.
Moving forward with your annotations is a straightforward process with HyperLabel. All of the tools you need to manage any labeling project are right at your fingertips, to handle, modify or edit your images. HyperLabel is easy and intuitive to use, even for non-engineers.
If an image from your downloaded library is unusable (which often happens with images obtained via search results), you can simply skip it and move on. When you’ve completed labeling, the HyperLabel Dashboard will provide information about your new dataset, including how many photos you labeled and skipped, how long the process took, and more.