Audio classify model maker not work.

[SunJenry]

Have I written custom code (as opposed to using a stock example script provided in MediaPipe)

No

OS Platform and Distribution

Linux 20.04/MacOS

Python Version

3.8/3.10

MediaPipe Model Maker version

0.4.3

Task name (e.g. Image classification, Gesture recognition etc.)

Audio classification

Describe the actual behavior

Identify "up" and "down" with different language.

Describe the expected behaviour

Identify "up" and "down" with different language by a trained model from model maker.

Standalone code/steps you may have used to try to get what you need

https://colab.research.google.com/github/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/models/modify/model_maker/audio_classification.ipynb

Other info / Complete Logs

The colab provide by https://ai.google.dev/edge/litert/libraries/modify/audio_classification run failed. I tried with trained with Tensorflow to get a .keras model, converted to ".tflite" model.
Then loaded by MediaPipe Android, got error message: "Error: INVALID_ARGUMENT: Only 1D and 2D tensors are supported.; Initialize was not ok; StartGraph failed".

[UnityApex]

Is it usable on a Google pixel 8?

…

On Mon, Jun 30, 2025, 10:33 a.m. SunJenry ***@***.***> wrote: *SunJenry* created an issue (google-ai-edge/mediapipe#6027) <#6027> Have I written custom code (as opposed to using a stock example script provided in MediaPipe) No OS Platform and Distribution Linux 20.04/MacOS Python Version 3.8/3.10 MediaPipe Model Maker version 0.4.3 Task name (e.g. Image classification, Gesture recognition etc.) Audio classification Describe the actual behavior Identify "up" and "down" with different language. Describe the expected behaviour Identify Standalone code/steps you may have used to try to get what you need https://colab.research.google.com/github/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/models/modify/model_maker/audio_classification.ipynb Other info / Complete Logs The colab provide by https://ai.google.dev/edge/litert/libraries/modify/audio_classification run failed. I tried with trained with Tensorflow to get a .keras model, converted to ".tflite" model. Then loaded by MediaPipe Android, got error message: "Error: INVALID_ARGUMENT: Only 1D and 2D tensors are supported.; Initialize was not ok; StartGraph failed". — Reply to this email directly, view it on GitHub <#6027>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BTF6V5NA6N5RUBQVUXRZD633GFDFDAVCNFSM6AAAAACAONI5Y6VHI2DSMVQWIX3LMV43ASLTON2WKOZTGE4DQNZSGEZDENA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[SunJenry]

Is it usable on a Google pixel 8?
…

Filed to train custom audio classify model with official python code sample. Because I had not got my custom model file successfully, so I could not run with my custom model in Android.

The only success model I can use is official "yamnet.tflite".

My problem is I can not train my custom audio classify model with official python code sample.

[SunJenry]

https://ai.google.dev/edge/litert/models/modify/model_maker/speech_recognition, this link return 404.