Dataloop Ltd., a Tel Aviv-based startup that helps companies create training datasets for their machine learning projects, today announced that it has closed a $11 million funding round led by Amiti Ventures.

F2 Venture Capital, OurCrowd, NextLeap Ventures and SeedIL Ventures chipped in as well. Dataloop will use the capital to expand its presence in the U.S and Europe.

Before an AI model can be deployed in production, it needs to be trained on a large amount of sample data to bring its accuracy up to an acceptable standard. A neural network designed to analyze crop health, for example, would need to review thousands of aerial field photos or more to learn how to identify patterns of interest reliably. 

The most-time intensive task involved in putting together training datasets often isn’t gathering the files but rather preparing them. Every image, video or other record used for training needs to be annotated with contextual pointers, such as circles drawn around objects of interest in a photo, to ensure that the AI will reach correct conclusions while it’s learning. It’s this task that Dataloop focuses on simplifying.

The startup offers a platform that enables companies to upload their raw training data to a cloud-based environment and use machine learning to automatically perform annotation. Dataloop can, for example, add labels to traffic camera footage to describe the types of vehicles shown in a given frame. The startup says its service lends itself to use cases in other areas as well, including retail and agriculture, among others.

Dataloop also provides tools for performing data annotation manually, an approach that is time-consuming but is often favored by companies because it allows for a high degree of accuracy. The startup’s service has a drag-and-drop editor that workers can use to highlight objects or areas of interest in images. If needed, the automated and manual annotation features can be combined: A company could use machine learning to perform the initial labeling, then have its human annotators review the preprocessed files for accuracy.

Dataloop says it’s addressing a key technical obstacle in AI projects. “Many organizations continue to struggle with moving their AI and ML projects into production as a result of data labeling limitations and a lack of real time validation that can only be achieved with human input into the system,” said Dataloop Chief Executive Officer Eran Shlomo.

Before the $11 million funding round announced today, Dataloop raised a $5 million seed investment. 

Photo: Dataloop