Hello all, I have a question regarding the detection for a person/car with Code Project.
Right now if the person is within 10 feet of the camera then it get detected within BI but anything further it doesn't.
I did an experiment and I saved a video with a person and a car when they were about 30 feet away from camera and then I saved a screenshot from that video into an jpeg file. I then imported the jpeg into the Code Project AI explorer.
Both the car and person where detected there, however they were not detected within BI when it was part of the video that was getting recorded.
Does anyone have any idea what may be going on?

It's all about the settings:

1. If you have a fairly standard CPAI set up, it will look at a set number of images. If it finds something before that set number it will stop. It's possible the latest versions of BI5 don't do this now - check the Help file. You can also increase the number of images.
2. You can test the video in BI5, but it isn't the same as the still shots BI5 sent to CPAI.
3. The best way to see what is happening is to enable DAT files. Then you can see which model CPAI is using and what CPAI did with each of the images it was sent. You have to stop DAT files when you are done else they will fill up your folders.

Thank you.

I have been able to Ctrl+double click on the recorded alert to see the .dat file and what happens with the person detection, but how can I see why I get no detection if the image if a little further from the camera. I was hoping to go through the same process, scroll to the time in the video when the person should have been detected and find out from the dat file why it didn't happen, but it looks like the .dat file is created/appended only when detection happens.

The DAT file should be showing you multiple times and images as well as the data. You could test it by walking towards the camera from 30ft away and then view the DAT file.


Here:

1. I can click on different times, and it shows the corresponding image and adds squares around identified items.
2. It shows that I have "person,car,truck,bicycle" in "To confirm".
3. It shows I have "bus" in "Cancel" to make it check all images (This may no longer be required).
4. It has an asterisk next to the image that was determined to have the best id.
5. It shows that this camera has six post-trigger images.
6. It shows that I am using custom model "ipcam-combined" and that all other models are OFF.

It really should tell you what you need to know. This system checks an image every 100ms and uses a GTX1650 to minimise cpu load and accelerate CPAI identification.