To make it more interesting I thought it’d be cool to restrict the machine learning model to a single type of item and add support for voice control to change between types. While I was at it I added the audio from the Portal turrets, because why not?
Last night the object detection project, which is using a Raspberry Pi model 3b+ and an Intel / Movidius Neural Compute Stick 2 (NCS2), finally started working as expected. The video feed comes from a Pi cam v2.1. Kudos for Adrian at Pyimagesearch for his very helpful guide on this topic!