Hi! I'm a Computer Vision Engineer by trade, and I work in R&D and Product Development. I love working with the Oak 2 series of cameras since they load the AI compute onto the camera for execution, which saves you from using an external compute (therefore lower BOM cost, making your solutions more commercially viable). Sorry if this is the wrong spot for product feedback. Maybe this type of product/input is on your roadmap already.
Lately, on the side, I've been trying to get into embodied AI and play around with Robotic arm kits. These can be enabled with vision/CVML using a small camera. And most of these setups offload the AI compute to a Raspberry Pi 5 (weak AI performance) or a Jetson (which is too expensive for a hobbyist. Jetson nano super doesn't have great performance imo. These computes also handle some Ros 2 + Planning algorithms. For the controls, another MCU is used (mostly to run servos).
I know the Oak 4 has an MCU that can run Linux, but I'm unsure if it has the necessary I/O or hardware to drive the servo. However, if a potential future Oak series camera could incorporate Imaging, AI Compute, and a type of MCU capable of handling simple motion controls, servos, and connectivity. This could reduce the package of a hobbyist/student Embodied AI Robotics Kit to 3D-printed robot arms, servos, cabling, and an Oak camera.
I think embodied AI is going to be big, and vision is going to be a part of it. And I always recommend Oak cameras due to their philosophy of packaging everything needed to run vision applications into a single package. I definitely believe Luxonis can play a part in this space!
