Blueiris able to ignore blackout certain sections ONLY when sending to deepstack?

[joshwah]

Hello!

I am running BI + Deepstack and running it on high mode (GPU).

I am wanting to have certain regions completely ignored from deepstack analysis. However, I am not wanting to have those sections removed completely from the recordings... only the deepstack analysis.

Is this possible?
Thanks!

 

[joshwah]

Yes. If you edit Zone A and mask areas you want excluded from analysis, both, DS and BI will not include those areas in the analysis. You will still record the whole thing - everything that the camera sees.

Hmm. I've tried what you recommended however it appears it still analyses the whole image? For example see below screenshots:

P.S - i've got Zone A & B setup so when it crosses zone B it will trigger.

Zone A:


ZONE B:

 

[joshwah]

You might be using Zone crossings incorrectly. You should take a look at this YT video from HomeTech Video.
I don't know what you're trying to detect, so the link would probably be a good start.

It has always worked and still works. I just use a "single zone" in the object detection - see below screenshot.

The issue is, it still continues to look at the whole image instead of the masked parts.

Do you have an example of an image analysed where it shows it is obeying masked parts?

 

[joshwah]

Further, when i enable the option, "Blackout masked areas" it shows it like this... but as mentioned, it is still sending the entire image to deepstack?

 

[sebastiantombs]

DS will identify what it sees in a scene. The blackout areas are not sent to DS in the photos it analyzes. DS will trigger on what it sees in a scene. I have used the blackout areas in the camera, itself, to exclude areas from DS with success.

 

[wittaj]

Keep in mind that the DeepStack analysis during playback operates different than the live stream. The playback will show every item that DS recognizes.

This needs to be done for when someone is troubleshooting.

The only way to not have the entire image sent to DS is to have the camera itself black out the image that is being fed to BI, which I do not think you want.

DS is sitting local to your computer and not going out for analysis, so what difference does it make if it sends the whole image? During live view, it is only looking for what you have identified you want triggered on....

 

[sebastiantombs]

I get occasional triggers of masked areas if the detection fits the criteria entered in the detection objects list.

 

[joshwah]

Ah I see my issue... i had the option "testing & tuning --> anaylze with deepstack" enabled which always showed the entire image without the mask.. as soon as i disabled that option it showed the mask successfully!

 

[sebastiantombs]

Nope, I regularly get IDs of a dog that are shadows moving on a large tree trunk that is masked in BI motion detection. It only happens at one point in the day, when the sun is in just the right position but it does happen almost daily when it's sunny. I expect that to end when the leaves fall, say October/November.

 

[sebastiantombs]

I get real time, "branded", alerts that are "verified" as a dog. I haven't bothered to look at the .dat files because the alerts directory is a total mess and grew to almost 30,000 files between .dat and jpg files. I'm just getting that back under control and will re-enable saving the data for that camera, and one other, that produce those falses.

 

[JL-F1]

RE: .dat files
I notice they don't get deleted when I delete the alert in BI, I have to go to the folder directly and delete.
Shouldn't they be deleted when you delete the alert in BI