Figure Eight’s human-in-the-loop machine learning platform creates high-quality, structured training data for your machine learning models.
You upload a .csv of unstructured data into our platform, either via our secure web portal or RESTful API. We support text, image, URL, audio, or video data for myriad use cases. Once you upload your data, you can select exactly how you want it annotated.
Creating a Figure Eight job is simple. You can use one of our ready-made templates or deeply customize your workflow with our code editor. This step allows you to provide specific instructions about how you want your data annotated, customizable Java and formatting, conditional logic, and a whole lot more.
You set the rules for your Figure Eight job. You can build custom workflows, set your own accuracy threshold, segment for language or geography, and optimize your job settings for speed and cost.
We take quality seriously. Our platform offers test questions that make annotators understand exactly what you want before they’re allowed to label your data. We maintain trust scores for every person who tasks on our platform. And we employ redundancy and answer aggregations to ensure your data is labeled as accurately as possible.
We offer a real-time monitor page that shows you how your data labeling is progressing as well as reports on test question performance, contributor metrics, and more.
Whether it’s our judgment aggregation algorithm or our Machine Learning assisted Video Object Tracking solution, we bake machine learning into our platform, augmenting the best of human intelligence with state-of-the-art models.
Once your data is annotated, you’ll be able to download a full or aggregated report containing every label on your entire dataset.