Saving JPG on motion trigger

Post Reply
IAmATeaf
Posts: 48
Joined: Mon Jun 17, 2019 7:48 pm

Saving JPG on motion trigger

Post by IAmATeaf » Wed Jun 24, 2020 4:02 pm

I’m testing AITools with DeepquestAI and have configured a clone camera to save a JPG on motion. The clone I have has a main stream and substream set, thought I’d use the benefit of having a substream being monitored for motion.

The problem I’m having is that sometime the pictures are in the resolution of the main stream, other time it’s the resolution of the sub stream. Using masks in AITools this stops the mask being applied as it expects the mask to be at the same resolution as the JPG.

Anybody seen this or know if there might be an option that I can try? I could obviously get rid of the substream but would rather avoid having to do this as I like the CPU load that using sub streams gives.
User avatar
terk
Posts: 75
Joined: Tue Jun 18, 2019 2:36 pm

Re: Saving JPG on motion trigger

Post by terk » Wed Jun 24, 2020 6:08 pm

Gentle Pumpkin states in his instructions for AI.Tools that using a main stream for the normal camera and a substream for the cloned (hidden) camera actually takes more resources than just using the main stream for both, so I haven't messed with the substream on mine and have had no issues.
sparkzter
Posts: 35
Joined: Fri May 15, 2020 9:53 am

Re: Saving JPG on motion trigger

Post by sparkzter » Wed Jun 24, 2020 7:03 pm

Same for me...i use mainstream for both the 'hidden' AI cam and the normal stream (that is displayed and is recording 24/7)..i didnt have any increase in CPU load when i duplicated all of my cameras with an 'AI' twin.

AI cam settings (in the 'recording' tab) i have jpg ticked and quality for me i have set to 30% - processing the AI image at full resolution loads up the CPU and takes longer, ive found 30% is plenty good enough for accurate AI detection.
IAmATeaf
Posts: 48
Joined: Mon Jun 17, 2019 7:48 pm

Re: Saving JPG on motion trigger

Post by IAmATeaf » Wed Jun 24, 2020 9:07 pm

That is how I had it setup before but thought I could reduce CPU usage further by detecting motion on a low res substream thinking that BI would save the image using the main stream.

I’ll revert back as an image from the substream doesn’t have the clarity and misses some object detection.
Post Reply