cancel
Showing results for 
Search instead for 
Did you mean: 

transfer learning with fp ai vision

yessine
Associate III

so im trying to create a model that distinguish different types of pasta based on an already existng model(people detection in fp ai vision) . i followed the provided  guide step by step

https://wiki.stmicroelectronics.cn/stm32mcu/wiki/AI:How_to_use_transfer_learning_to_perform_image_classification_on_STM32?fbclid=IwZXh0bgNhZW0CMTAAAR25n0jxpKo1k2Uj9cMePZSWMndI6WrCJdcJVDusByLj9vjdIXR160Wz2X4_aem_429JIxKjhUbaTfukJ86K7w

i donwloaded the model and generated the files with x cube ai and cubeMX and mooved the files to the fp ai vision people detection project....

 

when i build the project the file network.c generated 110 errors and all errors are from it alone. the other files dont have a single error .and all the errors are likeScreenshot 2024-07-24 133417.pngexample of lines that have errors in network.c file(there is plenty like this in many other places in the filesexample of lines that have errors in network.c file(there is plenty like this in many other places in the files this type

Description Resource Path Location Type 'ai_layer_conv2d' {aka 'struct ai_layer_conv2d_'} has no member named 'in_ch_format' network.c /STM32H747I_DISCO_PersonDetect_Google_CM7/Application line 4063 C/C++ Problem

 

so is there any solution or guide you can help me with

3 REPLIES 3
fauvarque.daniel
ST Employee

Please see answer in this thread: https://community.st.com/t5/edge-ai/error-in-file-network-c-from-quot-how-to-use-transfer-learning/td-p/701946


In order to give better visibility on the answered topics, please click on 'Accept as Solution' on the reply which solved your issue or answered your question.
yessine
Associate III

*the other thread is from my mentor

i tried with x-cube-ai version v7.2  the errors of network.c have dissapeared and the project compiled properly . the problem now is when i run the project  . its doesnt run properly . the model get stuck in 55% (name)=(whatever i putted in the place of "not person from the original file of people detection" in  fp-ai-vision-app.c  line 125 

so is there anything else i need to do and not mentioned in the guide to make the model run properlyScreenshot 2024-07-26 121640.png

Hello,

For any new Computer Vision development on STM32, we strongly recommend to start from the application examples which are available in the STM32 AI Model Zoo.

The STM32 AI Model Zoo is available at the following link:

https://github.com/STMicroelectronics/stm32ai-modelzoo.

 

Thank you.

Julien.