Embedding Tensorflow Model In Firmware - Could Not Find File

I have built a custom model that loads and runs from the SD card. Now looking to embed the model into the firmware. I have followed these two posts, and placed my model “blob_classify” and label files into openmv/src/lib/tflm/models, then build the firmware.

I load the firmware with this path:
\wsl.localhost\Ubuntu\home\username\openmv\src\build\bin\firmware.bin

After loading the firmware I get the error “Could not find the file”.

When I change to model filename to “fomo_face_detection”, the model loads.

Hi, can you provide your model and labels? I just tried this feature out by adding a renamed second copy of fomo to the nicla firmware and there were no issues.

I’m building the firmware for RT1062. In OpenMV IDE the board is “OpenMV Cam RT1062 Legacy”.

Build using this:
make -j$(nproc) TARGET=OPENMV_RT1060 -C src

Figured it out. After adding the .tflite and .txt files to /model folder, need to force make to rebuild with -B switch

make -j$(nproc) -C src/micropython/mpy-cross -B

Great!

Hi @ajacobs,

I’m encountering the [same issue] (Error loading embedded model) you described regarding loading an embedded model in OpenMV. I was wondering if you could share more details about how you executed your model and the steps you followed.
Thank you.

It may be a memory issue. Try removing all other models from the folder.

1 Like