Sighthound Video 6 - Open Beta

April 30, 2019

Hello Sighthound Video Users -

Sighthound Video 6 (SV6) is now available as an open BETA!  You can download it from the Sighthound account portal.

As our CEO Jonathan mentioned in his previous post, SV6 uses our new Sighthound IO (SIO) computer vision platform.  Thanks to SIO, SV6 can detect:

  • People

  • Animals

  • Vehicles

Sighthound IO is also the foundation for other new features we will add this year.

We’ve been running this BETA for several weeks on our systems. Overall it works very well during the day - and has much improved accuracy across the new detection types.

Unfortunately it still isn’t great at night.  SIO still makes too many mistakes and often returns the wrong results in adverse (low or infrared) lighting.  

So far we’ve received nearly 20 terabytes of adverse lighting video from Sighthound Video users.  Thank you - we really appreciate it!

We are now annotating and training on this adverse lighting video and will soon release updates to improve our recognition accuracy in these conditions.

SV6 also makes it easy to submit your “problem” video clips to us to further improve such accuracy.  The full details of how we use these video clips is documented in our privacy policy; but the TLDR; version is:

  1. We will keep your video clips secure,

  2. We won’t share your video clips with any 3rd parties,

  3. We will only use your video clips to improve our accuracy.

The desktop, web, and mobile apps make it easy to submit problem clips.

In the desktop app you right click to send it to us, and can optionally provide a note about the problem:

In our iOS application just swipe left on a clip to submit it.  On Android, just use a long-press:

In our web interface you will find a “Send to Sighthound” button in the clip player view:

We also have many other improvements in SV6:

  • Bulk exporting of all clips within a time range

  • Export a clip as a GIF

  • Export a single-frame JPEG image

  • Control the size and encoding quality of exports

  • Buffer incoming video to RAM instead of to the system/boot drive.

We are pushing out updated iOS and Android applications that support detecting animals and vehicles.  These should be in the Apple app store and Google play store within the next few days. In the meantime the old apps will work with 6.0 but without the new detection types.

We have a few known issues we’re still working on:

  1. Mobile push notification clips are not correctly loading in the mobile apps

  2. We have a slow memory leak (approx 100MB per week)

  3. We are still addressing usability issues

  4. SV6 uses about 30% more CPU than SV5

  5. We are close to adding acceleration via Intel GPU’s

  6. We hope to add acceleration via NVidia GPU’s in a later 6.x release

I’m sure we will find other problems as more of you start using SV6.  We will jump on new problems quickly.

I want to thank every one of you that provided feedback to us on the direction of Sighthound Video - and for sticking with us these last few years. There is still much for us to do.  We will be more open and transparent with all of you. We have other improvements coming to Sighthound Video we hope to share soon.

Best regards,

 

RJ Burnham

President and CTO

rj@sighthound.com