Deepstack and not flagging some clips

[lcam]

I have DS setup with an additional 2 images to be analyzed and triggers setup with no zone crossing and Edge vector.
I will always get a trigger and a clip but sometimes DS will not flag the image with a person at my door. If I test run the clip through DS it will detect the person above the minimum confidence percent so I'm not sure why I don't get a flagged image. I have tried more images and less images to be analyzed but so far I get misses.

 

[aesterling]

I mentioned this in another thread, but have you looked at which frames BI is sending to DS for processing?

I had a similar issue where the first frame BI sent to DS was the leading-edge of motion and didn't contain enough of a subject for DS to identify, and then when the 2nd image was sent to DS one second later, the object was already out of frame.

 

[Coldair]

I mentioned this in another thread, but have you looked at which frames BI is sending to DS for processing?

I had a similar issue where the first frame BI sent to DS was the leading-edge of motion and didn't contain enough of a subject for DS to identify, and then when the 2nd image was sent to DS one second later, the object was already out of frame.

where do you see these images? new to deep stack and would much rather pay someone to set it up but no one is around here that wants to work

 

[lcam]

I mentioned this in another thread, but have you looked at which frames BI is sending to DS for processing?

I had a similar issue where the first frame BI sent to DS was the leading-edge of motion and didn't contain enough of a subject for DS to identify, and then when the 2nd image was sent to DS one second later, the object was already out of frame.

I have tried up to 4 additional images at 1 second intervals. I can try more I guess and see what happens. I'm not sure of the pitfalls of too many images.

 

[lcam]

I mentioned this in another thread, but have you looked at which frames BI is sending to DS for processing?

I had a similar issue where the first frame BI sent to DS was the leading-edge of motion and didn't contain enough of a subject for DS to identify, and then when the 2nd image was sent to DS one second later, the object was already out of frame.

I read the thread but I don't get the blue border as others are saying.

 

[aesterling]

It sounds like it's hit-or-miss for some users. It wouldn't hurt to email BI support to ask about it. Maybe there are additional requirements I'm unaware of.

 

[Coldair]

not seeing the blue border either

 

[lcam]

I'm pretty sure that when I run the clip through the DS test it is not using the same settings as I have for the triggers. Otherwise the results would be the same. As it is the test run with DS always detects a person whereas during the event the person may not be detected. Also when running the test DS detects objects that are not in my confirm list so that also leads me to believe the test uses different settings.

 

[tward392]

I'm pretty sure that when I run the clip through the DS test it is not using the same settings as I have for the triggers. Otherwise the results would be the same. As it is the test run with DS always detects a person whereas during the event the person may not be detected. Also when running the test DS detects objects that are not in my confirm list so that also leads me to believe the test uses different settings.

Same has been happening for me, just figured he would get the kinks worked out eventually!

 

[looney2ns]

If you don't email support, he'll never know about the issues.

 

[Coldair]

I sent ken an email but I bet he thinks i am an Idiot as I hab bI on a 8 tb drive instead of C: drive which I just upgraded to a 2tb SSD. anyway I was having a lot of strange errors but most of them stopped when I reinstalled BI and deep stack. now even my help file works

 

[lcam]

I found a setting that is helping DS detect the objects by, I think, delaying sending frames to it. I am using triggers only so on the trigger>AI page I unchecked the "Begin analysis with motion-leading image". I assume this now waits until the pre-trigger has passed before it starts sending images to DS. My pre-trigger is only 4 seconds but this seems to have helped successful detection's a lot.

 

[sebastiantombs]

I took the other tact and reduced the pre-trigger time down to .1 or .2 depending on the specifics of the camera location and what I'm trying to detect.