So, in other words, I just need to go back into each cameras settings and change my fps to about what I want it to be in BI. I knew that already, but thought maybe I could get away with it. That's what I get for thinking and not listening. What happens, if you set your fps higher in the cams by a lot? Just curious. Thanks. I'll go change my settings now.
To deepstack, or not to deepstack, that is the question??
- Thread starter ingeborgdot
- Start date
I also have to go look up about main stream and sub stream. They video I just watched talks about keeping your main stream and substream at the same fps. Is that actually right? I thought I saw something on that years ago, but have not thought much about it since.
looney2ns
IPCT Contributor
If you set the cams FPS higher in BI than what is actually set in the camera, you are wasting memory.
Sub Stream Guide
.substreamGuide td { border: 1px solid #000000; } .substreamGuide a { color: #0080ff; } .substreamGuide code.inline { margin: 0; background: #f6f6f6; border: 1px solid #e0e0e0; border-radius: 0px; padding-top: 1px; padding-right: 3px; padding-bottom: 1px...
Thanks looney2ns. I had just read that in numerous articles. I'll go back in and change all my substreams to a higher fps. I just need to start reading the help file in BI instead of just searching google. I find it in google, it's just that it takes me so much longer to find. I gotta break the habit. You are a fine person looney. Your picture drives me nuts, but I guess it goes with the name. 
The problem with that philosophy for one of my cameras, is that it won't allow me to go any higher than 15 for my substream. As I stated above, I have to have a higher fps so I don't get sick when I check over the video.
I may just have to get rid of it, although it does just fine I guess. It was my first camera. A Hikvision DS-2CD2342WD-I.
I may just have to get rid of it, although it does just fine I guess. It was my first camera. A Hikvision DS-2CD2342WD-I.
Okay, I'm sure it's in the reading somewhere and I'll go to BI first to read, but I have almost 200 alerts since I first started several hours ago. One of my problems is that there is a car that is parked across the street that is just in sight of the camera. It keeps saying truck, and 53% but it keeps putting alerts into the alert folder. What do I do to stop that?
I'll read and try to find out, but in the mean time, I'm just letting you know.
I'll read and try to find out, but in the mean time, I'm just letting you know.
The weakness of DS. DS analyzes a still photo for the objects in the photo, based on a motion alert from either BI or the camera fed to BI. If it sees a truck, it confirms the alert. I don't know of a good way to avoid that. For my applications, I don't have anything in my camera views, just yard. I'm not sure if there is a way to blank out part of the image that DS analyzes or not.
I went from medium to high for sensitivity. It said my door was a train. lol. That car was a truck, but now it says a car. I hope I can find how to fix this.
sebastiantombs
Known around here
kklee
Pulling my weight
DS is far from perfect, I have a covered BBQ on my deck that DS keeps insisting is a person. I masked out part of it in the detection zone and DS now ignores it. The static object detection for some reason doesn't filter it out, maybe because the cover moves a bit due to wind.
Sometimes the DS results are pretty funny, my wife and I have occasionally been detected as birds, cats, dogs, and bears. Probably due to camera angles and what we were wearing at that time.
Sometimes the DS results are pretty funny, my wife and I have occasionally been detected as birds, cats, dogs, and bears. Probably due to camera angles and what we were wearing at that time.
I have Detect/ignore static objects checked. Is that the one you mean? What about apply to motion triggers only??? That one was unchecked. I checked it. I'll go read up on how to mask in that area.
Last edited:
sebastiantombs
Known around here
DS does seem to have problems with stationery objects. My pick-up and the wifes' car are both IDed constantly at night.
I fixed part of the problem. I just set 2 of my cameras differently. I made them so they don't hit the street anymore. I'll see how I like this, and if it doesn't get any better, I may just go back to the old way. It seemed to be more efficient, at least the way I had it setup. There needs to be something that takes care of the stationary objects better. Maybe there is and I have not READ about it. 
Interesting to read about the static objects challenges, not having any issues with cars on the driveway but they are close to the camera and maybe this makes a difference. Generally the DeepStack is working well especially now I trained a custom model for night schedule. Looking at the computer spec of @ingeborgdot then I would think you could crank the settings fairly high on DeepStack, assuming the GTX 1660 is doing the work, increase the '+ real-time images' and reduce the 'analyze time' to 250ms or 500ms. I've done the same on my critical cameras and it manages but the CPU is struggling, got a GTX 1660 on order and hope that will take the strain.
Increasing the camera frame rate I am sure has also helped with my general DeepStack accuracy, this may be due to my CPU which is only an i5-4690 but the system has 16GB RAM so shouldn't be too sluggish. I also tested the DeepStack on another computer running i5-8400 with 32GB RAM and the CPU struggled on that also without pushing too hard, both these systems are running high end Samsung SSD.
Increasing the camera frame rate I am sure has also helped with my general DeepStack accuracy, this may be due to my CPU which is only an i5-4690 but the system has 16GB RAM so shouldn't be too sluggish. I also tested the DeepStack on another computer running i5-8400 with 32GB RAM and the CPU struggled on that also without pushing too hard, both these systems are running high end Samsung SSD.
Last edited:
Oh, the specs for the computer is for my main office computer. My BI server has no graphics card in it.
Ah, apologies, my mistake
- might still be worth trying the above, I'm currently only on CPU for DeepStack and they seemed to help, of course it can be very scene specific. Certain settings will help one setup and hinder another.
Last edited:
Is DS completely local and works without internet? i read it needs an API key but hoping thats a one time thing during setup.
sebastiantombs
Known around here
I thought it did with one of the tutorials i've read or it need the key to download, but its its 100% local then will for sure look into it.