Longer AI analysis?

iKu777

n3wb
Joined
Jan 25, 2023
Messages
4
Reaction score
0
Location
Canada
Hi,

I have a camera that is set with a motion detection and Codeproject AI to detect cars coming into my driveway.
During the day it works fine but I have a problem at night.
At night incoming cars shine lights in the drive way which triggers the object detection way earlier than the car makes it to the motion detection zone. Looking at the canceled alert video feed, the camera is analyzed for 5 seconds and the AI doesnt find any object (which is accurate in this case).
But I would like the camera to keep analyzing longer (let's say for 10 seconds) so that the car has the time to get into the zone and be detected as a car.
I have tried various settings but not sure how to make the analyzing time longer.
Anyone could help with that?

Thanks
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,153
Reaction score
49,019
Location
USA
Add DoNotCancel into the to cancel field and then it will run all the additional photos you assign.

But every field of view is different and you will need to figure out the combination of number of additional images and time between each one in order to get it to identify the vehicle.
 

iKu777

n3wb
Joined
Jan 25, 2023
Messages
4
Reaction score
0
Location
Canada
Increasing the total of "real-time images" * 'analyze one each' would give me the total time that is covered by the analysis, is it?
For example, if the motion was triggered 5 seconds before the car arrives in the zone and my settings were 10 images, 1 every 500ms, is would be expected that it didnt get detected, am I correct?
 

iKu777

n3wb
Joined
Jan 25, 2023
Messages
4
Reaction score
0
Location
Canada
And the 'DoNotCancel' setting would only help if it detected something before and stopped detecting.
But it my case where nothing was detected, that setting wouldnt help, correct?
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,153
Reaction score
49,019
Location
USA
The DoNotCancel means it will run AI on EVERY additional image you put in. So if you put down 10 additional images, it will run AI on all 10. Without DoNotCancel, the first amount of light shine would then come back as nothing detected.

So if your field of view has light shine for 5 seconds before a vehicle comes into view and you tell it to run 4 additional images at 1 second each, it will miss it as the car doesn't come into view until the 5th second. Tell it to do 10 images at 1 second intervals and it will return back a car then.

But it is a balance - vehicles at different speeds and different types of headlights and alignments will cause the number of seconds of light shine to be different, but you don't want to tell it to run 200 additional images either.
 
Top