New Blueiris camera box suggestions

Dave Lonsdale

Pulling my weight
Joined
Dec 3, 2015
Messages
456
Reaction score
195
Location
Congleton Edge, UK
With the discovery that the 100/170 and other errors when using DeepStack with Blue Iris are likely to be CPU overload, I would suggest that PC recommendations are put on hold until after the final chapter. And delete the related wiki however that’s done.
 

sebastiantombs

Known around here
Joined
Dec 28, 2019
Messages
11,503
Reaction score
27,709
Location
New Jersey
I think it would be better to add some cautions to the Wiki regarding BI/DeepStack. My experience, so far, has been that with an NVidia GPU doing the DeepStack work, things work pretty well. Maybe newer CPUs, gen 9 and 10, may work well also, at least from what I've seen.
 
Joined
May 1, 2019
Messages
2,217
Reaction score
3,511
Location
Reno, NV
I think it would be better to add some cautions to the Wiki regarding BI/DeepStack. My experience, so far, has been that with an NVidia GPU doing the DeepStack work, things work pretty well. Maybe newer CPUs, gen 9 and 10, may work well also, at least from what I've seen.
back when I was in the midst of my tire slashing issue and initially learning Deepstack, I snagged a Nvidia GPU at a local auction for $10. It's no super duper gaming video card. But somewhere I read the CUDA value of the card is what matters. Days later, I thankfully learned about using Dahua AI so never got around to installing this card. What difference on the CPU did your card make?
 

sebastiantombs

Known around here
Joined
Dec 28, 2019
Messages
11,503
Reaction score
27,709
Location
New Jersey
@Holbs It made a very big difference. The only version of DS I could run, well the latest version anyway, was 5.4.8.2. It worked OK but still missed an awful lot of obvious detections. Switching to the GPU version lowered CPU utilization about 5% and eliminated the spikes when DS ran python for detection. The detection times have dropped from anything between 500 and >1500ms on the CPU to as little as 55ms on the GPU. The detection rate is virtually 100% on all but one camera and that applies only to night when headlight bloom blinds it. Yes, it does take some additional power to run the NVidia card, in my case about 100 watts, but it is a small sacrifice, and worthwhile to me, to get good performance from DeepStack. I now have confidence in it when it comes to verified alerts versus missed alerts.

Installation does take a little time but is well worth the effort as well.
 
Top