Blue Iris - Occupied? What am I doing wrong?

[Vansmak]

Disclosure: I'm VERY to Blue Iris and have been reading a ton of threads trying to fine-tune my setup...While I getting close, I still have time where BI sends me an 'occupied' alert instead and what I think is a pretty easily identified object (in this case a car, see clip below). My setup is a i7-6700 with 8 gig ram, running a SSD Hard Drive as the main boot drive and a 2TB platter as my storage drive. I have three cameras currently - two Loryta turrents and an Amcrest Doorbell AD410. The PC is dedicated to Blue Iris running Windows 10 and recording direct to disk on alerts only. My cams are all set to 15fps as well as the sub stream and iframe. I also lowered the bitrate to 1024 and have enable Hardware decoding Intel + VPP.



These are my trigger settings:



Everyone was so helpful in getting me with the right cameras on this forum....now I just a little help getting them to record correctly! Thanks in advance!

Also:: I have no idea why the video stutters at the end when I export it.....It plays fine inside Blue Iris prior to export.

 

[wittaj]

Night time is always more difficult. It is obvious it is a car to you and I, but if that was not the type of image a model was trained for (headlight glare), then it is only seeing your parked car and calling it occupied.

Uncheck the use main stream if available because that is only causing time delays - the first thing Deepstack does is downrez it to analyze it, so it serves no real purpose.

Add a few more images - go from three to 10 as the headlight glare/bounce blows thru the 3 images before the car is in view.

Add something in the cancel like banana so that it forces it to look at 10 images and get a better chance of seeing a good side profile of the moving car.

Check the Save the AI analysis so that you can see what is happening.

Consider making one of the zones for the street and tell it to only analyze that zone (but only do this if the above doesn't work).