# Hardware and Software Setup

We recommend the following setup in order to run the examples described in this guide or to develop your own Qt application for object detection.

## Hardware
- Ethernet camera connected to the display

- Google Coral mPCIe AI hardware accelerator board installed into the display

- Linux image including drivers for the mPCIe chip on the display

How you connect and integrate a camera to the display is described in the documentation for the `Camera Module`.

The Linux image that includes drivers for the mPCIe board needs to be installed onto the display. It can be requested from us.

💡 It is possible to run the object detection using an external USB dongle version of the Google Coral accelerator, for example on a V700 display. This concept is however mainly designed for experimental and development use, and not for commercial use. Futhermore, it is using USB 2.0 and thus has a lower transfer speed than the mPCIe board.

![Coral mPCIe Board](images/coral_mpcieboard_75.png)
![Coral USB Dongle](images/coral_usb_50.png)

<span style="font-size:75%">Figure 1. The mPCIe and USB Dongle versions of the Google Coral device.</span></br>

## Software
- Edge TPU driver

- Tensorflow Lite library

- The gstreamer element qmlglsink

- OpenCV

You will need the software mentioned above both for building the applicantions in Qt and for running them on the display. All the software packages can be requested from us and are very easy to install.

![Setup](images/ai_setup.png)
<br><span style="font-size:75%">Figure 2. Example of a setup for running object detection with a display.</span></br>

## Troubleshooting
If you have problems to run the object detection on the display, you can first verify that the Linux image contains the correct drivers and that the mPCIe chip is accessible. It can be done with the following commands in the terminal:

Check loaded kernel modules. It should return something like this:
~~~
$ lsmod | grep apex
apex                   24576  0
gasket                 94208  1 apex
~~~

Check the PCI bus, which should look something like this:
~~~
$ lspci -nn | grep 089a
01:00.0 Class 0880: 1ac1:089a
~~~

Check that the mPCIe board is available as a Linux device:
~~~
$ ls /dev/apex_0
/dev/apex_0
~~~

You can also run the diagnostic messages tool *dmesg* to see if it reveals any information about the issue.