Enough is enough
Moderator: Staff
5 posts
• Page 1 of 1
-
JesseS - Posts: 31
- Joined: Sat Dec 30, 2017 9:54 pm
Enough is enough
Have been using this for a couple of years and still lots of issues. Running on a dedicated iMAC quad core i5 and it Boggs down at 9 cameras with 100% cpu use, all cameras shut off at random times and I have to reboot to get them back. 'People' detection is a joke as almost every car at night or day triggers it. This is a VERY resource intensive program and still needs a lot of work, I have invested $350 in this and am going to cut my loses and start looking for another solution. Just the fact that all of my cameras go down when I am 500 miles away is reason enough not to trust it. I have connected to my cameras both manually and with ONVIF and it makes no difference.
-
ZrC - Posts: 9
- Joined: Thu Jun 01, 2017 11:54 am
Re: Enough is enough
Before anyone can assist:
-What year is the iMac?
-What camera resolution and frame rate?
-Camera connection to Sighthound via ONVIF or RTSP?
I was running on an old second gen low voltage quad core Xeon and it handled 7 cameras (1080p 10fps) easily with only 1-2% usage for each on the cpu.
-What year is the iMac?
-What camera resolution and frame rate?
-Camera connection to Sighthound via ONVIF or RTSP?
I was running on an old second gen low voltage quad core Xeon and it handled 7 cameras (1080p 10fps) easily with only 1-2% usage for each on the cpu.
-
cwizard - Posts: 30
- Joined: Sat Dec 16, 2017 11:37 am
Re: Enough is enough
Intel Xeon CPU here, 3ghz or something. 13% utilization with 5 cameras up and a RUST server going. I'm going to say an iMAC may not be a suitable host.
-
mr.Magoo - Posts: 29
- Joined: Wed Jan 23, 2019 8:35 am
Re: Enough is enough
These threads are always painful to watch as there's typically two camps, the bashers and the lovers.
The only reason I post here is that I feel I'm in no mans land, somewhere in between.
Motion detection is going to be resource intensive no matter what, a 1080p camera @ 6fps and a 4K/8MP @ 30fps will put a completely different load on the system so a mine v.s. yours comparison doesn't really work unless you have all the details.
Sighthound is MUCH less resource intensive compared to some of the alternatives when using the same PC and the same cameras, again, this is on a Windows Server, no GPU support. It might be that the Mac's don't run as efficient but mine (3.6ghz E3 Xeon) was using about 5-6% per camera / process, so 9 cameras plus another process for the main process would result in roughly 60% utilization on my box with cameras of the same type that I've been using (5-10mp @ 7fps, which is low fps).
I DO agree that the detection engine need some work and I've asked for this before - calibration.
Unless the algorithm cleverly figures it out by itself, but I don't see a reference to size/distance.
Not even Axis does it automatically: https://www.youtube.com/watch?v=hghAtNd6vAQ
I bought, loved and stopped using Sighthound all within a year and there's a few things that drove that cycle.
I had really good cameras but very bad/basic internal motion detection so I was evaluating pretty much all options out there and I found SightHound... it was love at first sight, did everything I wanted, great.
Bought a $250 pro license (and 6 months later they introduced the home version which would have been perfect for me) and I was very happy... for a while...
I live on a fairly quiet street but my logs where still cluttered with temporary videos from cars driving by and reviewing past recordings are slow and painful at best. So at this point I resorted to a somewhat unorthodox method and that was to make Sighthound a "plugin" to another NVR as I liked the motion detection features but didn't care at all for the NVR. So I had SH trigger webhooks to a NodeRed raspberry PI which in turn triggered a motion event to the NVR software I was using, benefit of doing this is that I could (A - make custom events to the NVR software and B- I could tweak the timestamps as there's a ~5s delay between the event occurring and the webhook fire).
This worked just fine and I was kind of sort of happy. That's until we had an episode with a underage house-party down the street and a bunch of drunk kids running around, puking on the sidewalk, hit 'n run when leaving the party and minor brawls. (again, this is normally a VERY quiet street).
At this point I realized that despite 5/10mp resolution the cameras aren't all that great (they're still considered "professional").
I needed (well, mostly wanted) PTZ and Auto Tracking.
While I swore I would never buy a certain brand due to the controversy surrounding them, I swallowed my "pride and did it anyway, and I could not be happier (for the money).
With these new cameras came better (built-in) motion detection and with that the need for Sighthound disappeared... I wish I would have known this a year ago and saved $250, but I like to tinker, so.. oh well..
@Steve Guerra (Not sure if this works)
I've been trying to figure out who SH's target audience is and maybe I should have read up on the financial reports first. If Sighthound IO is the main revenue stream and Sighthound Video is just a way to debug / test algorithms then I get it, SH Video is always taking the back seat in terms of development and features.
But if SH Video is it then, like I said, who is the target audience ?
I doubt professional users (i.e. mid-to-large corporations) would use it since the review process is so tedious.
The majority of home users will use whatever box came with the cameras and/or exists with their NAS.
That leaves the hobbyists / enthusiasts and in the past I think many of them would have been scared off by the high price tag of a pro license (much better with the new home license) and gone with a less expensive / or even free option.
That ultimately leaves users like who I was... enthusiast who's frustrated with the lack of features in their current cameras... can't be many of us around
If I may suggest one thing... Well two then...
1. Make a genuine plugin for a certain NVR software (I think you know which one I mean), you'd still retain the home users/customers since it's free (the NVR), but now you also have a chance to attract businesses/pro users. Should also allow you to focus on the AI rather than NVR portion. I for one would buy a new license just for that plugin.
2. If you're set on keeping the NVR functionality, please improve the review process and how you handle recordings. Also allow us to mask of areas of no interest to reduce processing and "false" alarms.
The only reason I post here is that I feel I'm in no mans land, somewhere in between.
Motion detection is going to be resource intensive no matter what, a 1080p camera @ 6fps and a 4K/8MP @ 30fps will put a completely different load on the system so a mine v.s. yours comparison doesn't really work unless you have all the details.
Sighthound is MUCH less resource intensive compared to some of the alternatives when using the same PC and the same cameras, again, this is on a Windows Server, no GPU support. It might be that the Mac's don't run as efficient but mine (3.6ghz E3 Xeon) was using about 5-6% per camera / process, so 9 cameras plus another process for the main process would result in roughly 60% utilization on my box with cameras of the same type that I've been using (5-10mp @ 7fps, which is low fps).
I DO agree that the detection engine need some work and I've asked for this before - calibration.
Unless the algorithm cleverly figures it out by itself, but I don't see a reference to size/distance.
Not even Axis does it automatically: https://www.youtube.com/watch?v=hghAtNd6vAQ
I bought, loved and stopped using Sighthound all within a year and there's a few things that drove that cycle.
I had really good cameras but very bad/basic internal motion detection so I was evaluating pretty much all options out there and I found SightHound... it was love at first sight, did everything I wanted, great.
Bought a $250 pro license (and 6 months later they introduced the home version which would have been perfect for me) and I was very happy... for a while...
I live on a fairly quiet street but my logs where still cluttered with temporary videos from cars driving by and reviewing past recordings are slow and painful at best. So at this point I resorted to a somewhat unorthodox method and that was to make Sighthound a "plugin" to another NVR as I liked the motion detection features but didn't care at all for the NVR. So I had SH trigger webhooks to a NodeRed raspberry PI which in turn triggered a motion event to the NVR software I was using, benefit of doing this is that I could (A - make custom events to the NVR software and B- I could tweak the timestamps as there's a ~5s delay between the event occurring and the webhook fire).
This worked just fine and I was kind of sort of happy. That's until we had an episode with a underage house-party down the street and a bunch of drunk kids running around, puking on the sidewalk, hit 'n run when leaving the party and minor brawls. (again, this is normally a VERY quiet street).
At this point I realized that despite 5/10mp resolution the cameras aren't all that great (they're still considered "professional").
I needed (well, mostly wanted) PTZ and Auto Tracking.
While I swore I would never buy a certain brand due to the controversy surrounding them, I swallowed my "pride and did it anyway, and I could not be happier (for the money).
With these new cameras came better (built-in) motion detection and with that the need for Sighthound disappeared... I wish I would have known this a year ago and saved $250, but I like to tinker, so.. oh well..
@Steve Guerra (Not sure if this works)
I've been trying to figure out who SH's target audience is and maybe I should have read up on the financial reports first. If Sighthound IO is the main revenue stream and Sighthound Video is just a way to debug / test algorithms then I get it, SH Video is always taking the back seat in terms of development and features.
But if SH Video is it then, like I said, who is the target audience ?
I doubt professional users (i.e. mid-to-large corporations) would use it since the review process is so tedious.
The majority of home users will use whatever box came with the cameras and/or exists with their NAS.
That leaves the hobbyists / enthusiasts and in the past I think many of them would have been scared off by the high price tag of a pro license (much better with the new home license) and gone with a less expensive / or even free option.
That ultimately leaves users like who I was... enthusiast who's frustrated with the lack of features in their current cameras... can't be many of us around
If I may suggest one thing... Well two then...
1. Make a genuine plugin for a certain NVR software (I think you know which one I mean), you'd still retain the home users/customers since it's free (the NVR), but now you also have a chance to attract businesses/pro users. Should also allow you to focus on the AI rather than NVR portion. I for one would buy a new license just for that plugin.
2. If you're set on keeping the NVR functionality, please improve the review process and how you handle recordings. Also allow us to mask of areas of no interest to reduce processing and "false" alarms.
-
JesseS - Posts: 31
- Joined: Sat Dec 30, 2017 9:54 pm
Re: Enough is enough
I think my issue it the graphics card, it is a older AMD Radeon HD 6750M 512 MB card, I don't think it can handle the video bandwidth. I really don't want to drop a bunch of money on a new computer. I am testing Xeoma and it handles all 9 camera's live preview with a 35% CPU load without issue, but then it does not record all camera's 24/7 like SHV, only on motion detection with pre & post record. I got an extended testing license for free so will run it thru its paces, seems to have many more settings and options than SHV, with a steeper learning curve.