How about you remove the cutom models from the folder you don't need. I've always done that. That should work, correct?
5.8.8 - March 4, 2024
- Thread starter fenderman
- Start date
I wish Ken would fully clear this up.
I have most cameras using "ipcam-general" , some using "ipcam-combined", one using "license-plate" and one using "delivery". I just want to make sure that they aren't going to start using all 4 on all cameras since I've specified the model I want to use for each camera. So is this behavior changed now, and if so why?
Some cameras stop as soon as they've found an object, others keep going because I have some non-existent object specified in the "to cancel" field. Again, is this behavior changed and if so why?
I don't usually update unless it's necessary, so I'm definitely holding off. But at some point if I do, I need to know if/how this is changing.
looney2ns
IPCT Contributor
Can't answer the "why" part, but this change occured with 5.8.7.8, see below:
5.8.7.8
Improvements have been made to the heuristics used when deciding which image among all analyzed to use for final alert confirmation or cancellation. It is no longer necessary to place a “dummy” object into the the “to cancel” box for processing to continue regardless of a single confirmed object.
Can't answer the "why" part, but this change occured with 5.8.7.8, see below:
5.8.7.8
Improvements have been made to the heuristics used when deciding which image among all analyzed to use for final alert confirmation or cancellation. It is no longer necessary to place a “dummy” object into the the “to cancel” box for processing to continue regardless of a single confirmed object.ipcamtalk.com
Sadly this feature didn't improve my image for selection and I need to keep a dummy object in the to cancel.
Hmm...I'm surprised that is still working: from the sounds of that 5.8.7.8 write-up it shouldn't actually add any additional analysis by doing so. Strange that you can still "force" the added images.
Perhaps what happens in your case is that the "best" image occurs towards the end of a long string of consecutive analyses.
Yeah it is one I use to get a car in the center of the alert image at night. Without it I continue to get just the front or back of the car and sometimes just barely the head or tail lights LOL
Well, decided to give 5.8.8.8 a go since there has been no update for a few days. So far 15 hrs in and not one error and no restarts of BI service. Also, My AI settings are the same as they have been for a year or so and it is working perfect. Not sure if this was in help file before or not but Ken added a few examples.
So on my settings below it is working as expected.
So on my settings below it is working as expected.
Last edited:
Same problem with 5.8.8.latest. First time the entire PC locked up and I had to force a power off and restart. Seven and nine days later the PC restarted after running out of virtual memory - somewhere around 80 GB. Both time the committed memory was less than 12 GB and holding steady the night before. Looking at the log files a bunch of processes could not obtain more memory. When this started happening BI was using ~75GB. within an hour BI was using an additional 4-5GB. Very soon after the second warning message the PC restarted.
I have switched back to 5.8.7.11. Too soon to tell is this fixes the problem. Going to be away for a few weeks so I have no choice but to hope so.