Back from 2012 / Start SV at System Startup? / Cats,Dogs...

[B00ze]

Good day.

Well after using VitaminD (for free) for 5 years, I finally moved the program to a more powerful computer with a bigger disk (went from 36GB to 320GB). But I increased camera resolution, so now I can store LESS video than before, lol. I could still run this for free, but I am buying a license now; I might want some of the extra features and 5 years for free is a good run.

I see some of my "I want that" have made it into Sighthound (no more using NSSM to run as a service + we now have international date/times formats). This is great! I see Ryan is still around; Hi There! I also see that Sighthound, despite a complete re-write, still flags cats, dogs and light reflections as people. If this was a neural-net I'd say train it more, but I forget what the technology is, I don't know if it can be trained...

Ok, my question: What does the "Start Sighthound Video at system startup" option do exactly? Since Sighthound installs a "starter service" which runs at system startup, I fail to understand what the extra option in settings does (I haven't really tried to figure it out, someone please let me be lazy and explain?)

Thank you.
Best Regards,

[Kbond]

Hello,

So, the "Start Sighthound Video at system startup" option is exactly what you are thinking it is. Just another option to start Sighthound on start up. You can read more about it below:

https://www.sighthound.com/support/runn ... r-computer

Hope that helps. Please let us know if you have any other questions or concerns, thank you!

Regards,
-Kevin

[ryan]

Hi!

[B00ze]

Hi Kbond, Ryan.

I was asking because I noticed that the Sighthound service starts regardless of the state of the Start At System Startup option. So I did a little testing, and if the option is not selected then video recording does not start upon reboot; the service reads that option and acts accordingly.

About detection: I browsed your website, and you use deep (neural) networks. Possibly this means the output of one network is fed into another, with several layers before the final product. This keeps each network small, but maybe it's time to make the networks Sighthound Video uses a little bigger? The processing power Sighthound Video requires is really low; maybe there is room to increase processing requirements in order to improve detection of humans (vs cats and dogs). Just a thought.

Thank you.
Best Regards,

[ryan]

You're correct that the toggle controls whether the app is launched on startup or not, while the service itself is ever-present. It needs to be there to respond to user app launch requests and such as needed but it is using no resources while idle so is not a drain on your system by being installed and enabled.

Bringing additional algorithms into SV is something we're looking forward to doing as we're able. It's great to hear preemptive feedback that you'd be willing to take a performance hit to get them.

[B00ze]

Hey Ryan.

Yeah, I noticed the service needs no resources at all; I was just trying to see how it all worked - Having come from VitaminD, I was curious how Sighthound did the service thing (I had been using NSSM with VitaminD.)

As for detection, I just finished reading your release history and I see that you guys are already improving the algorithm/networks (e.g. Version 4) as you see opportunities. I had browsed the forums yesterday and seen people saying cats and dogs detected as people and I knew from my experience with VitaminD that this happened. But what I had/have not done is give version 5 a chance; I just installed it 2 days ago (and bought my shiny new license yesterday); maybe it will not see my cats as people anymore? I'll report back in a few months if I remember to do so.

I'd like to say I am impressed with Sighthound. When VitaminD became Sighthound I thought that the party would be over, but I see that you seem to be in very good health, with new products and corporate customers. The Sighthound Video product has evolved quite a lot; there is even an Android App! I am happy to be one of your customers

Best Regards,

[B00ze]

Good day.

Well, I've had the new SV for weeks now, and can report back on detection. It very rarely detects cats as people now, this is much better than it used to be back in the VD days. However, it is horrible at detecting people, so much so that I have edited my rules to detect "any object" instead. I have the camera pointing straight at the door and while it almost always detects me as a human coming in (it really picks-up the door opening at first, and calls that a human, but it eventually moves its focus to me) it almost never figures I'm a human when I am leaving through the door. This might be because I turn on the lights before leaving, and it gets confused for a while, I will have to do some more testing. But it's possible it is too demanding now in determining what is a human...

Regards,