BlueIris Codeproject.AI not alerting at night

[kujo2023]

I cannot figure out why I am not getting confirmed alerts at night on my Reolink RLC-81ma camera. The Reolink camera software detects the car and records the event every time. However, BlueIris has been missing it every night. In BlueIris when rerun the clip with "Testing & Tuning" "Analyze with AI" turned on I can see that it clearly detects the car (see attached image). So why is it not showing up in my confirmed alerts list? Do I have a setting selected that is causing issues? I have included my nighttime settings for this camera. Also, what does "Occupied" mean in the snapshot list?

Another issue I sometimes have with this camera, which is a dual lens camera, is that only one stream will trigger an Alert but the other wont. I would like to have it setup such that if one triggers and alerts both will alert so that I will have both the wide angle and zoom angles alerts. I tried doing this with "Trigger Camera Groups", but it did not work for me. Any idea how I force that to happen?

Thanks for your help in advance.

 

[wittaj]

Keep in mind that the "Analyze with Deepstack/CodeProject" under "Testing & Tuning" will ALWAYS perform better than live as it is after the fact and should not be used as an analysis tool to try to figure out why it didn't see and trigger for a car or person. It should only be used to see what AI can find in that clip, like "hmm I wonder if AI can find a toothbrush" and then walk around with a toothbrush and have it identify it. I can run this on a camera not using AI and it will show EVERYTHING that AI has in its objects to find that it sees in the clip. This method will show you EVERY ITEM AI searches for.

You need to review the .dat files as that will show you how CodeProject interpreted it.

And keep in mind the Reolink can be problematic with BI. At night under the BI camera status page, what is the KEY number? It should be 1.0 but no less than 0.5. Reolinks have been known to drop to 0.1 and that is problematic for motion detection in BI.

 

[kujo2023]

Keep in mind that the "Analyze with Deepstack/CodeProject" under "Testing & Tuning" will ALWAYS perform better than live as it is after the fact and should not be used as an analysis tool to try to figure out why it didn't see and trigger for a car or person. It should only be used to see what AI can find in that clip, like "hmm I wonder if AI can find a toothbrush" and then walk around with a toothbrush and have it identify it. I can run this on a camera not using AI and it will show EVERYTHING that AI has in its objects to find that it sees in the clip. This method will show you EVERY ITEM AI searches for.

You need to review the .dat files as that will show you how CodeProject interpreted it.

And keep in mind the Reolink can be problematic with BI. At night under the BI camera status page, what is the KEY number? It should be 1.0 but no less than 0.5. Reolinks have been known to drop to 0.1 and that is problematic for motion detection in BI.

Ok I checked the .dat files for the timeframe of the car's movements. What I found was that the car moved (with AI detection) between 4:10:58 and 4:11:06. The .dat files were not active during this interval. I have .dat files for 4:10:31 and 4:11:26 for both of the streams (wide & zoom) from the camera but both missed the movement of the car. I think I need to force the generation of more .dat files. What setting do I need to change to not miss the this? I have tried setting the AI "real-time images" to 6 and the" analyze one each" to 500ms but this didn't help. I am not sure what setting to tweak. It is my trigger settings? I really am not sure where to go on this.

 

[wittaj]

As I mentioned, what is the KEY number in the BI camera status page like this one. As you can see, the ones where KEY is below 0.5 will be problematic and can miss a trigger. A key of 0.10 means if an object is in and out of the field of view in under 10 seconds, BI might not trigger for motion.

You need to confirm your KEY value is 1.00 but no less than 0.50. If it is less than 0.5 that is the problem. If it is above it, then it is settings or trying to do too much with one field of view.

 

[Kujo]

My Key values are 1.0 for the wide angle lens and .5 for the zoom lens. So that does seem to be the issue.

 

[Gunn]

Did you ever figure out the proper settings to get this camera and Blue Iris CodeProject.AI to alert on the car at night? I am having the very same issue, with a similar Reolink camera. My "key" value is also 1.0.
Motion triggers on the car approaching with the bright headlights, but CodeProject.AI doesn't find anything. I can only assume it's due to the bright lights where the car behind them isn't "seen" properly. But can't figure out what to tune to get it to reliably work?

 

[wittaj]

Did you ever figure out the proper settings to get this camera and Blue Iris CodeProject.AI to alert on the car at night? I am having the very same issue, with a similar Reolink camera. My "key" value is also 1.0.
Motion triggers on the car approaching with the bright headlights, but CodeProject.AI doesn't find anything. I can only assume it's due to the bright lights where the car behind them isn't "seen" properly. But can't figure out what to tune to get it to reliably work?

Post screenshots/video of your problem field of view.

It may simply be the camera cannot get a clean enough picture either because the firmware likes to play with your settings regardless of what you set, or you are on default/auto and need to manually set a faster shutter.

Then you need to add an item in the TO CANCEL box that would be an item you wouldn't expect to see.

That will force it to use EVERY additional real time image. If you don't put something in the too cancel box, then as soon as AI thinks it found or didn't find something, it cancels out.

Here is an example. At night the camera AI would struggle with this tight view. It has a straight on angle of the street to get a side profile of a car and it would miss it a lot of times because the vehicle just isn't in the field of view long enough, so this is a great candidate for CPAI.

Now the issue I had with CPAI is that it would either find a car but the alert image would be the lightshine on the street or just a part of the vehicle, or it would trigger out nothing found due to headlight bounce off the street.

CPAI has a "to cancel" option, which means it will analyze EVERY image to determine if the item is in it. Once I added a cancel Giraffe in the field, it now will go thru all the images and select the best one, which gives me the whole vehicle in the frame and it eliminated the nothing found as well. It makes for scrubbing video much quicker as I can skip looking at video of known vehicles.