So I finally got the model working for the motion recognition using data from X-NUCLEO-IKS01A2 on NUCLEO-L476RG. I downloaded the generated tensorflow lite library but when I try to build the example that @janjongboom uploaded to github (https://github.com/edgeimpulse/example-standalone-inferencing-mbed) to test the model directly on the board, I get the following error:
Error: Conflicting types for ‘arm_abs_f32’ in “edge-impulse-sdk/CMSIS-DSP/Source/BasicMathFunctions/arm_abs_f32.c” in the mbed online compiler.
I think that’s a compatibility error between the CMSIS-DSP library and the mbed os library but I can’t find a way to solve it .
@doliprane Looks like CMSIS-DSP is included twice. If you add the following to your
.mbedignore would it continue?
@janjongboom No, when I do that I get an undefined function error. Apparently, as I looked for a solution on the net, I found out that there is a problem between mbed library and CMSIS-DSP library when using mbed online compiler. I’m going to try and export the project to an offline editor. Maybe that can solve the problem.
Hmm… could be. We have only tested this with GCC 9 using Mbed CLI, not with the Mbed online compiler. Please let me know what you find!
@janjongboom So I switched to Mbed CLI and it worked just fine. Apparently it’s a problem with the online compiler. Thank you