Coming soon: Custom activity regions

You asked and we listened. In addition to exciting announcements from TechCrunch Disrupt NY around Dropcam Tabs and people detection, we’re also looking forward to launching a popularly requested feature: customizable activity regions. With activity regions, you’ll be able to select and label specific areas in your camera’s view, and set alerts and other preferences based on activity within those regions.

Traditional approaches to motion detection work by detecting differences between video frames. This can lead to false events, where shadows, light reflections or other unimportant changes can trigger a detection. While it may seem simple enough to reduce unwanted alerts by setting a region of interest, many of these unwanted lighting changes can still enter into those regions.

In order to make activity regions work like you’d expect, we had to refine our motion system further. When you look at a video, your brain is doing something much more complex. It is tracking moving objects and ignoring things like shadows or reflections. You can probably do this without even thinking about it — but for computers it’s an incredibly difficult task. To enable something like activity regions, we had to figure out a way to teach Dropcam how to track objects.

What this all means for you is that you’ll be able to receive alerts only when someone or something actually enters the areas you’ve set up. We think it’s going to be popular and can’t wait to get it into your hands later this summer.

Tweet about this on TwitterShare on FacebookShare on Google+Share on LinkedIn

4 thoughts on “Coming soon: Custom activity regions

  1. This sounds great, I have been waiting for a feature like this since the dropcam first came out. We have a few trees/bushes that sway in the wind and we get constant motion events all day. Its hard to see when people walk by because if the bushes/trees are moving around the same time it ignores these events.

    Really excited to try this out.

  2. Pretty cool . I constantly get false alerts for headlights of cars passing by, etc. This should help prevent that kind of thing. Also, hopefully I can help the algorithm know where my kitchen is vs my entry. Right now it does an OK job. It has figured out 2 activities on its own and I have labeled them “Kitchen” and “Entry/Hall” (the latter is my front door entry and the hallway). These two activities work OK, but I am not sure what region or types of movements it is picking up; hopefully Custom Activity Regions will help.
    Also, what would be nice would be if Dropcam told you what movement it is detecting/where in the video. Even better would be the ability to DropCam’s algorithm that this type of movement is irrelevant (i.e. a shadow, car headlights passing by, movement of a tree, etc) so you can help the algorithm learn. If DropCam setup a feedback section or check list that allowed the users to identify why this movement is not relevant (i.e. check here if movement is from a tree or bush, shadow, change in light in the room, etc) this would be a free source of data to train their algorithm and DropCam might end up with the best video recognition algorithms out there. This is similar to what Google did with its 1800 GOOG 411 and voice recognition.

  3. Jason, you said when something ‘enters’ the region. What happens when something moves ‘within’ the region, or when something ‘enters’ the region? By the way Jason, this needs to be an ‘activity’ monitor. I hope that is what you mean. Perhaps for light changes you only need a threshold of change.

  4. What happens when something leaves the area? You need to stop playing, and become pragmatic in your approach. You other ideas might be a nice weekend hobby for you.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>