I am trying to build an app with Kivy which takes an image and returns the super-resolved version of it. In order to do so, I am using a pre-trained super-resolution model. It is working well on my computer and I can build it to apk file using buildozer.

However, when I try to install it on my android device it crashes and gives the following error:

ImportError: dlopen failed: "/data/data/.test..../files/app/_python_bundle/site-packages/tflite_runtime/_pywrap_tensorflow_interpreter_wrapper.so" is for EM_X86_64 (62) instead of EM_AARCH64 (183)

Before implementing the app with pre-trained model, I built a simple app using PIL library which takes an input image turns the grayscaled version of it and it worked fine both on my computer and my android device.

Then, I built the app with the pre-trained model and I could never achieve to run it on my android device. First, I built it with the latest versions of each libraries (i.e. python3==3.11, tflite-runtime==2.14.0) and it did not work.

I searched for the error and saw this. After that I followed the instructions here. I checked the recipe page and saw that tflite-runtime recipe works for version 2.8.0. I changed my requirement versions to match that (i.e. hostpython3==3.9.21, python3==3.9.21,kivy,pillow,numpy, tflite_runtime==2.8.0, pybind11==2.9.1). I still get the same error.

I also changed the recipe provided under p4a to match the newest version (2.14.0) of tflite-runtime, which did not work either.

I think I am missing something here, but I cannot see it.