Voxel51, an open-source software company based in Ann Arbor, Michigan, has developed a commercial service called FiftyOne Teams, which assists computer vision AI models in inferring the correct object.
Voxel51's cofounder and CEO, Jason Corso, lives at the University of Michigan, where he obtains his PhD as a professor, and where he got the idea to start the new company. In recent years, computer vision adoption has increased, as has the size of the datasets.
Corso said of VentureBeat, "I had datasets that were tens of thousands of samples and I could look at every one." "Now my students have come along and they can't look at a million samples, and so the need for Voxel51 has arisen."
Today, the business announced that it has raised $12.5 million in series A funding from Drive Capital, Top Harvest, and Shasta Ventures, as well as existing investors eLab Ventures and ID Ventures, and the University of Michigan.
The potential and limitations of unstructured data for computer vision
Unstructured data may take many forms and include any type of data that does not fit in a specific data structure format (e.g., columns and rows).
Video content is increasing exponentially as the number of cameras continues to rise internationally. There are programs that allow users to look for images taken at a certain location.
Although there is no shortage of unstructured image data and large datasets used to train computer vision models, ensuring accuracy is a challenge.
"Our whole philosophy is that when datasets grew to be over ten million samples, nobody bothered to examine the images anymore," Corso said.
Voxel51 serves as a bridge between what a data engineer does when creating datasets and what that same engineer or their partner does when they're training models. It can be used to identify potential errors as well as to compare the performance of different models.
Voxel51 allows users to semantically slice data to determine whether or not a model is correct. For instance, a Python API allows a user to search through a computer vision dataset to discover all of the images in which one model outperforms another, for images where a child is running into the street.
The enterprise and open source
Voxel51 began as an open-source software, but now it is officially launching its FiftyOne Teams enterprise offering, which includes commercial support and additional capabilities.
The Voxel51 open-source project was first launched in August of 2020 and has grown over the previous two years, with up to 150,000 monthly users. “The open-source project is built for a user with local data, where all data is on a single system,” Corso said.
FiftyOne Teams, a commercially supported service, offers cloud data as well as role-based access control (RBAC) to enable many users to use the same platform securely. Currently, the commercial service is not offered as a fully managed cloud service, instead organizations will need to run the technology on-premises or in their own cloud instances.
“We are envisioning a future in which, at least for certain types of customers, perhaps startups who don’t want to go and deploy locally into their ecosystem, a managed service, but that won’t be released for some time,” Corso said.