A collaborative comparison chart for Vision Processing Devices. To note:
- This is not meant to be exhaustive but rather highlight the sensors and devices that come across my inbox/social media feed. It's actually impossible to keep track of all the "Intelligent" Cameras out there. There are way too many cropping up way too quickly, especially small ESP-32-like devices that can do face recognition, or run a reduced version of ImageNet which recognizes 5 or 10 objects. One defining characteristic of this product vertical is how easy it can be for fabs and design houses to iterate out new, often purpose-specific cameras and processors.
Rather, I'm trying to keep track of devices and systems that come across my notice that have some significant features or differentiator from others out there. THere is, for example, an interesting split in verticals between CPU-based programmable devices like the OpenMVCam and the Vector Processor-powered OAK series. EyeCloud dispenses with the notion of a default camera and instead gives you a CSI/MIPI port to bring your own(or buy one of theirs). For a GPU powered device, a camera is just another source of matrixed data. Being able to attach a FLIR sensor or even something more exotic like a scintillation counter to an "edge" device presents some interesting use cases.