Umbo CV raises $2.8M seed to create smart security cameras that prevent crimes

Umbo CV has raised a $2.8 million seed round for its security cameras, which use artificial intelligence to identify suspicious activity and prevent crimes before they happen. The Taipei- and San Francisco-based startup’s funding was led by AppWorks Ventures, with participation from Mesh Ventures, Wistron Corporation, and Phison Electronics.

The two-year-old startup has already shipped its system—including cameras and a cloud-based management platform—to clients in Dubai, the United States, and Europe, and will begin mass manufacturing next month. Co-founder and chief executive officer Shawn Guan says Umbo CV has also received $1.4 million in pre-orders and letters of intent from other customers. Its funding will be used for research and hiring.

Other members of Umbo CV’s founding team include chief technology officer Ping-Lin Chang and chief scientist Tingfan Wu, who both hold doctoral degrees in robotic vision. It hopes to add more artificial intelligence researchers to its team in Taipei.

Before launching Umbo CV in 2014, Guan spent more than seven years working in the surveillance technology industry. He tells TechCrunch that Umbo CV was created because he wanted to focus on something besides developing high-definition videos, which many companies use as their chief selling point.

“That doesn’t help customers solve problems in the security industry, because it can only give you high-quality video after the event has occurred. If someone gets hurt, we can show clients a better video, but that didn’t stop someone from being hurt,” he says.

“In the industry, I got frustrated with spending so much money in building infrastructure but not getting a lot out of it. People still get away with crimes, so we have to find a way to solve it.”

As Umbo CV grows, it will face competition from manufacturers such as Huawei and Seagate, which are working on smart security cameras, and systems like AISight. Umbo CV wants to differentiate by building algorithms that combine images from multiple cameras and scan them for anomalies (for example, a car driving against traffic or someone scaling a wall).

Guan claims that is more effective than the geometric models used by some rival AI systems. Geometry-based intelligence scan for things like someone holding up an object that may be a weapon—but, of course, can also just be a wallet or something equally innocuous.

By scanning for anomalies, Umbo CV’s AI “doesn’t necessarily know what kind of event it is, but it knows what is not normal so we can point it out to security officers,” says Guan.