Computer vision dev platform Roboflow raises $20M

Roboflow, a Des Moines, Iowa-based startup developing tools for building computer vision models, today announced that it raised $20 million in a series A round led by Craft Ventures. It brings the company’s total raised to $22.2 million to date, and CEO Joseph Nelson says it’ll be put toward ongoing product development and hiring efforts.

The global computer vision industry was estimated to be worth $11.32 billion in 2020, according to Grand View Research. But while the technology has obvious enterprise applications — computer vision algorithms can be trained to perform tasks like spotting gas leaks, counting pills, and monitoring workplaces to enforce social distancing — companies can face barriers to leveraging it in production. Teams are often faced with rebuilding software infrastructure and attracting the necessary machine learning expertise.

Founded in 2019, Roboflow is the brainchild of Brad Dwyer and Nelson, who see computer vision as a foundational technology that can enable developers to solve problems resulting from devices’ inability to see the world around us. The platform provides a framework for developers to build computer vision into their products, enabling them to upload images and videos to train custom or prebuilt computer vision models.

“Computer vision is one of those generational technologies that, like the personal computer or mobile phone itself, will be adopted by every industry. Software is limited by its ability to receive structured information as input, and that structure is typically provided by a person,” Nelson said in a statement. “What computer vision does, is enable every part of the world around us to become programmable, unleashing a Cambrian explosion of applications. That’s why computer vision needs to be a part of the toolkit of every developer, not only expert machine learning teams.”

Dev-focused platform

With Roboflow, customers can annotate images while assessing the quality of datasets to prepare them for training. (Most computer vision algorithms require labels that essentially “teach” the algorithm to classify objects, places, and people.) The platform lets developers experiment to generate new training data and see what configurations lead to improved model performance. Once training finishes, Roboflow can deploy the model to the cloud, edge, or browser and monitor the model for edge cases and degradation over time.

“The promise of Roboflow reminds me of the early days of Stripe,” Roboflow investor Lachy Groom said in a press release. “Like payments, computer vision is a critical piece of infrastructure that needs to be made broadly available to developers. Consider how FaceID aims to unlock phones seamlessly, or how mobile check deposit alleviates the need to wait in line at a bank, simultaneously allowing bankers to focus on customer service.”

Roboflow competes with CrowdAI and Chooch, among others, in the growing computer vision development tools market. But Roboflow claims to have over 50,000 users, including engineers in half of Fortune 100 companies, startups, universities, and hardware companies including Pfizer, Walmart, Amgen, and Cardinal.

“These [computer vision] examples are the tip of the iceberg. Every industry will be rewritten, and for that to happen, all developers — not just machine learning experts — need to have tools that make vision accessible,” Groom continued.