Android Question TensorFlow Lite and Yolov8

[daniele.verdilio@beghelli]

Dear,
I'm quite new in B4A application and I find wonderfull the post from Moster67 and Zhshazly on Tensorflow Lite.
Now I'm triing to test a Yolo V8 image recognition and I've converted to .tflite.
But when I start the application B4A I receive the following error:

" Caused by: java.lang.IllegalArgumentException: Internal error: Cannot create interpreter: Didn't find op for builtin opcode 'RESIZE_NEAREST_NEIGHBOR' version '3' "

Maybe that the .aar is very old and now is necessary to have a new wrapper for the lates version (tensorflow-lite-2.9.0.aar) that require also the new API dependence (tensorflow-lite-api-2.9.0.aar) ?

Is there some body that can help me ?

Thanks

Daniele

 

[pixet]

If you are just looking for a library update try to take a look here:

tensorflow/tensorflow-lite/2.16.1

TensorFlow Lite

 

[daniele.verdilio@beghelli]

Dear Pixet,

thank you for the suggestion. I've found the aar file from Maven repository but I do not know how to "wrap" inside a new Tensorflow (like the Moster and Zhshazly)
I've tried to simply rename the new version of .aar with the the old name "tensorflow-lite-2.2.0.aar" but it work till version 2.5, after the interface is changed and does not work.

My model come from a Yolo v8 implementation and it is not compatible with that version of aar and wrapper.

Is there somebody expert that can help me to find a way ? I'm implementing Tensorflow-lite in Android Studio example but B4 is million times better and simpler ....

Thank you

Daniele