5.4.8 - June 16, 2021 - A DeepStack status page has been added

[mostlychris]

The asterisk is a confirmation that the alert is either confirmed or cancelled--in this case cancelled.

In your example, when you scroll down through the analysis you'll likely not find 'person 66%' or any other object you've designated in the To Confirm box of the camera's AI settings. I'm guessing you've also designated a number of additional images to analyze at 750 msec? I can't explain how the person was missed with this setup, but ensuring your AB motion zone rules are adequate, reducing CPU usage while analysis is going on (select Low or Medium in the AI tab of BI options), and consider not using facial recognition, which as of now seems to cause more problems than it's worth.

Not sure what you mean by scrolling and not finding person 66%. It did find person in a number of images in the left box. I have it set to analyze 14 images@750ms. I wonder if that should be a different time. I also turned off facial rec for now. I don't use it and it doesn't really work anyway as far as I have been able to tell. My zone settings are full image in zone A and then in Zone B I set the area I am interested in per the docs and discussion from BI on setting this up. I only trigger on motion in Zone B. I've played with enabling and disabling "Apply to motion triggers only" and haven't seen much change. Any other settings to be concerned with?

 

[beepsilver]

No need to check apply to motion triggers only. If you're not using facial recognition, you can take unknown from the To Confirm box. 750 msec is fine since you're not looking for fast moving vehicles on that camera. You might consider lowering it if you want to detect bicycles flying by or fast runners

On that .dat file analysis you posted, start from the bottom of the analysis--that's where the very first image occurs....at T-0 or T-some msecs. You can influence this number with changing the leading image setting. If the alert was cancelled, then likely you won't see person 66% in the analysis you posted with the .dat file. The .dat file is what actually transpired which is far more reliable than the option under testing and tuning.

 

[mostlychris]

No need to check apply to motion triggers only. If you're not using facial recognition, you can take unknown from the To Confirm box. 750 msec is fine since you're not looking for fast moving vehicles on that camera. You might consider lowering it if you want to detect bicycles flying by or fast runners

On that .dat file analysis you posted, start from the bottom of the analysis--that's where the very first image occurs....at T-0 or T-some msecs. You can influence this number with changing the leading image setting. If the alert was cancelled, then likely you won't see person 66% in the analysis you posted with the .dat file. The .dat file is what actually transpired which is far more reliable than the option under testing and tuning.

Makes sense. FWIW, going from bottom to top, person is shown in a number of the images in that analysis. It just never alerted. I'll play around more and do some additional tweaking and watch it over time. Thanks for the responses.

As a side note and maybe related, I seem to have issues with alerts if a static object is in view--such as my truck. This info I got from Sam at BI.


If a car is driving by, BI motion trigger will activate -> Deepstack will respond with objects including your truck -> an alert will be sent IF the BI motion detection overlaps with one of the objects, i.e. the car on the road. I think we still annotate all the objects found.



So if a squirrel walks in front of the camera, your truck annotated will still be there (I think... need to confirm with engineering), BUT no alert will be sent because the squirrel motion does not overlap with your parked truck.

 

[beepsilver]

For any 'person' in a frame before the asterisk or with the asterisk, what sort of icon do you see next to 'person' in that analysis? If you see an icon of a person with a percentage above your specifications, you should probably contact support because the software isn't behaving properly and you likely aren't the only one experiencing this. If with 'person' you have a percentage at or above your specification, but the icon is a stopwatch, overlapping monitors, etc., then I'd take another look at how you have your motion zones setup. Now that you're using DS to confirm an alert to avoid false positives, you can afford to reduce minimum contrast and make time to get the camera triggered sooner in order to give DS as much time as possible to begin analysis. Each camera likely requires a different setup, so where objects move quickly, you'll need the most time and zone space.

 

[mostlychris]

For any 'person' in a frame before the asterisk or with the asterisk, what sort of icon do you see next to 'person' in that analysis? If you see an icon of a person with a percentage above your specifications, you should probably contact support because the software isn't behaving properly and you likely aren't the only one experiencing this. If with 'person' you have a percentage at or above your specification, but the icon is a stopwatch, overlapping monitors, etc., then I'd take another look at how you have your motion zones setup. Now that you're using DS to confirm an alert to avoid false positives, you can afford to reduce minimum contrast and make time to get the camera triggered sooner in order to give DS as much time as possible to begin analysis. Each camera likely requires a different setup, so where objects move quickly, you'll need the most time and zone space.

It's a person above my confidence percentage setting. Guess I'll contact support.

 

[beepsilver]

It's a person above my confidence percentage setting. Guess I'll contact support.

Mostlychris, forgot to add (and you may have read this elsewhere already), but if you modify custom model settings in BI, you should restart your computer to be sure all DS components have properly closed and new components are started anew. I do this whenever I update BI to a new version as well.