Dropcam gets $30M for expansion plans and computer vision R&D

In 2009 Aamir Virani and Greg Duffy raided hardware stores to find parts to build an easy home security camera for Duffy’s dad. His father wanted to catch a neighbor’s dog pooping on his lawn. Today, the resulting company Dropcam has raised $30 million in funding from led by Institutional Venture Partners.
Existing investors Accel Partners and Menlo Ventures participated as well as new investor Kleiner Perkins Caufield & Byers. This funding brings the total amount raised to $47.8 million, which is a lot of money until you hear Duffy explain that the company didn’t actually need or seek out the Series C round. Duffy doesn’t disclose financials, but does say that Dropcam’s revenue grew 500 percent year over year.
That explains why Kleiner and IVP came knocking. Dropcam, the HD wireless video camera that sells for $150 has been a pretty big hit on the consumer electronics front. And it has done this by selling its product online and in the U.S. only. Duffy explains that with the funding he now has the capital to focus on getting into mass market retailers which Duffy hopes will boost sales by another 500 percent or 1,000 percent.
Dropcam might also get to expand to the rest of the world. Boosting distribution and marketing are common uses of growth capital, but Duffy has more up his sleeve. Dropcam is a completely vertically integrated product. Like Apple, which make sit’s own iPhone chips, Dropcam makes the physical camera as well as providing developers with the protocols and codecs that compress the video feed into something more manageable over slower broadband connections.

Here’s another video company on Amazon Web Services

Here’s where I tell you that the Dropcam can send up to 60 gigabytes of data per month back up to the cloud for your later perusal. That’s a big burden on smaller pipes, which is why Duffy says Dropcam has worked so hard on its compression algorithms and it’s software clients. But, there’s more.
Dropcam hosts all of the data people send to its cloud (about 39 percent of Dropcam buyers also subscribe to additional cloud services that include archival access to footage), on Amazon Web Services. Duffy says that Dropcam ingests more video per minute than YouTube, but won’t say anything more other than his staff measures data in petabytes. Put another way, YouTube says it ingests 100 hours of video a minute.
That’s a lot of data, making Dropcam a potentially huge AWS customer. Duffy says the company has optimized a variety of AWS features for the video-specific tasks Dropcam does as part of that vertically integrated strategy. And soon it will announce more ways it is putting that data to use.

Dropcam and the search for computer vision

Duffy says that Dropcam hired a team of people a year and half ago to start analyzing the data it has with the idea of making computer vision algorithms that work. The idea being that computer vision would help people reduce the time their camera is running or help identify patterns that might benefit Dropcam or the consumer. Given how much data is locked inside surveillance cameras, finding better ways to train computers to see is a big deal that companies ranging from Google to IBM are trying to solve.
Duffy argues though that will all the inbound video data available to Dropcam, it will be able to make greater strides than others. Because when it comes to building algorithms more data is better.
Duffy says that in the coming year Dropcam will announce a framework for computer vision that it built using the petabytes of user footage it has. The trick, though is that it built this framework without letting its data scientists see the footage. To protect user privacy (a pretty big deal when we’re talking about cameras in people’s living rooms and bedrooms) the team working teaching computers to see was essentially blind.
When the framework is out, then Dropcam will develop products and services associated with computer vision — another reason it was happy to take the money. And depending on how effective those algorithms are, the resulting services might be a hugely compelling aspect of Dropcam’s business. Because while 70 percent of Dropcam’s revenue comes from hardware sales today with the remainder coming from services, it’s the services and the data that will drive success in the coming wave of the internet.