Poll: What has been your leading constraint when designing embedded vision applications?

Training/AI model

Usually it's the proprietary camera drivers and camera connector / non standard pinout for DVP or MIPI interfaces. IMX219 , OV6540 as an example.

Adding vision covers a large array of possible issues.

The requirements of each application can affect the hardware, software, camera selection, lenses, lighting, contrast, exposure times, pixel resolution, pixel sensitivity, the list goes on.

I have worked on those types of projects for over fifty years and counting.

While it may appear to be just a simple question, it is not.

DAB

I have no interest in embedded vision in any of the stuff I work on (sensors, analogue and digital control systems, test equipment),

so happily 100% problem free

MK

I voted Combination.

On the surface it is software - installation and setup is a struggle for me, its complicated

Second, while it is software - is training the AI

For us the biggest pain points are cables and connectors. We need thin cables fully shielded and fairly long. The connectors need to be very small, rugged, shielded and waterproof. And of course low-cost. Custom cables can cost more than the circuit card assemblies.