Closed rmknan closed 5 months ago
Hello @rmknan
I did not add any build system. Here, you have many ways of doing so. I will say two of them:
If you need any assistance, I could add the build system.
I followed through the 1st one and made a C++ project. I can build now but I have been facing issues while trying to include the various headers that one has to. I tried including the basic ones and the ones with TF but I face errors.
If you could answer both it would be great!
Thanks!
Hello again @rmknan
Ok,
Right click over the project and click properties
. There go to C/C++ Build
> Settings
There in Tool settings
click in MCU GCC Compiler
and in Include paths
and the following:
In MCU G++ Compiler
in Include paths
You shall select C++14.
In addition, In Preprocessor
you also need to define STM32F429I_DISC1
, STM32F429xx
, STM32
, STM32F429ZITx
, STM32F4
in both GCC and G++ Compiler
In my case, it is just that I don't like the idea of the STM32CubeIDE, it is more a preference. So, in my case, I use STM32CubeIDE for debug and develop of the app and I was expected to introduce an independent build system, but I just change to other project/ideas jeje
But If you face any major issues doing this, I could add the build system with stm32cube idea which I think that I could not spent to much time on doing so
I did include the files but I face these errors. These were the errors that i faced before too. These are some of them but all of them are centered around arm_nnfunction
or CMSIS_NN
. I tried including these directories in the Include paths
but nothing changes.
fatal error: arm_nnfunctions.h: No such file or directory fatal error: arm_nnsupportfunctions.h: No such file or directory fatal error: kiss_fft.h: No such file or directory make: *** [tensorflow-lite/third_party/cmsis_nn/Source/PoolingFunctions/subdir.mk:34: tensorflow-lite/third_party/cmsis_nn/Source/PoolingFunctions/arm_avgpool_get_buffer_sizes_s16.o] Error 1
Hello @rmknan
Which FW
version are you using?
I just recorded this small video to show you how I did this. This is on ubuntu, but I shall be similar in Windows.
https://drive.google.com/file/d/14G14hAFc-xnRKZxfebsdhDmlFrnan_E-/view?usp=sharing
Best regards,
Which
FW
version are you using?
STM32Cube FW_F4 V1.28.0 I am using Ubuntu 22.04 C++14 is selected in the G++ tool settings
Is that what you asked?
I did the same as done in the video but I face the errors mentioned above.
PS - I have not done any TF projects on STM before.. I have Bazel installed but maybe something else is needed?
Yeah, that's what I was asking for. I think that I see what is going on. It was my fault, sorry. I was testing an old version of the project.
You also need to include:
tensorflow-lite/third_party/kissfft
tensorflow-lite/third_party/cmsis_nn
tensorflow-lite/third_party/cmsis_nn/Include
In addition, you shall define:
In both: MCU GCC Compiler
and MCU G++ Compiler
MCU GCC Compiler
MCU G++ Compiler
PS: This does not have anything to do with Bazel. In general, the workflow working with tensorflow-lite
is just getting the source code from here. From there, you could generate the folder structure that you use in your project. For using that project, you will need certain dependencies. However, here I just did that job.
I hope that now you could use the project.
Great!! It works now. Thank you!! I also had to change the sysmem.c
file. It was using some old terminology.
I would like to learn more about TF in MCUs.. Could you tell me how you started with it?
When I write on the touch panel It seems I have to write it in a mirrored way, Where can I change that?
I think i also need to calibrate my panel, the numbers only appears when I write on the bottom half.
Hi,
Could you please elaborate what you mean with mirrored way?
About tf in mcus. I wrote an small entry about how I got into this small project. Basically, I enrolled into a edx free course. This gave the idea of how this work.
The documentation from Tensorflow is quite good. In addition, in the original repository there are some examples. At the moment that I did this project (sept-2020) there was quite few documentation and few examples. Now, this situation has been improved. There are examples which I think are interesting to grasp some ideas about how it works.
Could you please elaborate what you mean with mirrored way?
I cannot write normally as you would on paper. The directions are reversed. If I move my pencil up then it draws a line going down.
Thank you for the links I will look into it.
mmm, I'm not really sure, maybe it is realted with the calibration process because I have tested without any problem as you can see in the repo
Could you tell me how do I upload this to the MCU? The STM IDE does not recognize the project and am unable to build.
Thanks for the help.