Why am I getting compile errors when deploying a matlab function that uses snapshot() from raspberry pi cameraboard together with deep learning?

I have two sections in my function. First, it captures an image from the raspberry pi camera module. Deploying it on the hardware as standalone was successful. Then I added another set of code — using AlexNet to classify the captured image. After making the necessary changes in coder.DeepLearningConfig() (a new feature in MATLAB 2020a), this error came up during deployment:

error: cannot convert ‘uint64m_T*’ to ‘uint64_t’ for argument ‘1’ to ‘int EXT_CAMERABOARD_snapshot (uint64_t*, uint8_T8*, uint32_T*)’

.

.

.

recipe for target 'cameraboard.cpp.o' failed

It seems like there was a problem with compiling the cameraboard when using it with deep learning models. To check that there was really no problem with AlexNet, I tried deploying it without the cameraboard and it worked. Are there any fixes for this?

Answer 1

[Reeno Joseph]

Hello Miguel,

This looks like a bug in the cpp codegen for cameraboard. We will fix this issue ASAP.

Meanwhile you can use webcam() API as a workaround.

-Reeno

[Miguel Dorado]

I'm getting this error when running it in pi. I'm trying to change the default resolution so I used the sytanx wcam = webcam(mypi, cameraName, resolution). The funny thing is that even after I typed the exact cameraName, I would still get this error

[Miguel Dorado]

[Reeno Joseph]

Please try the index to access webcam.

raspiObj = raspi();
cam      = webcam(raspiObj,0);