Yes, the static buffer example - the one we have as default for all the supported boards is just meant for testing the inference on your training/test data on device and making sure the results match - and in your case they did, which is great (you ran inference on an image without bees and model detected nothing).
If you’re asking what do you need to do to run inference on the camera stream from ESP-EYE camera module, then you’ll need to write a code for obtaining the camera image from the sensor, processing it (resize and crop) and then passing onto run_classifier function of EI SDK. You’ll find some pointers in these two files
You’ll need quite a bit of C++ proficiency to put them together though. In future we plan to add support for ESP32 Arduino builds, but I cannot give a time estimate on that at the moment.