Is the object recognition running on the AI built into the camera, or on the development platform (PC, Embedded Linux devices etc)?

The recognition is using the 3D Object Detection API for ZED2 cameras. Also, at the moment, only Persons and Vehicles can be detected and tracked with this API. 

You can find more information about it in the link below :

You can also check our GitHub page:- so that you can look at the various use-cases and applications that we have developed and currently working on.



The ZED2 is a stereoscopic camera. This means that the camera outputs left and right synchronized images. 
Then the ZED SDK compatible with all ZED cameras provide features such as : 
- Real-time depth map at different FPS/Resolutions
- Real-time positional tracking
- Real-time SLAM (3D mapping)
- Real-time object detection and body tracking.

The ZED SDK only requires an NVIDIA GPU to be used since it's linked with CUDA dependency. It is also free of charge.