Hair segmentation with mediapipe on tensorflow lite

[Arezki93]

Hi,

I am trying to use mediapipe's hair segmentation model in script but I am coming across the error of

RuntimeError:`` Encountered unresolved custom op: MaxPoolingWithArgmax2D.Node number 12 (MaxPoolingWithArgmax2D) failed to prepare.

can you please explain how did you created the without custom op version of the mediapipe model ?

[sgowroji]

Hi @Arezki93, Could you please provide the steps to reproduce the above error. Thanks!

[Arezki93]

Hi @sgowroji, thank you for the answer, here are the steps I followed : 1) I loaded the hair_segmentation.tflite file 2) I did a allocate_tensors() -> it generates an error (Encountered unresolved custom op: MaxPoolingWithArgmax2D.Node number 12 (MaxPoolingWithArgmax2D) failed to prepare)

import tensorflow as tf import numpy as np interpreter = tf.lite.Interpreter('hair_segmentation.tflite') interpreter.allocate_tensors() **--------------------------------------------------------------------------- RuntimeError Traceback (most recent call last)

in 1 model_file = 'hair_segmentation.tflite' 2 interpreter = tf.lite.Interpreter(model_path = model_file) ----> 3 interpreter.allocate_tensors() ~\anaconda3\lib\site-packages\tensorflow\lite\python\interpreter.py in allocate_tensors(self) 406 def allocate_tensors(self): 407 self._ensure_safe() --> 408 return self._interpreter.AllocateTensors() 409 410 def _safe_to_run(self): RuntimeError: Encountered unresolved custom op: MaxPoolingWithArgmax2D.Node number 12 (MaxPoolingWithArgmax2D) failed to prepare.**

[google-ml-butler[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you.

[TomerJakobovich]

I also encountered this issue, @Arezki93 did you find a solution?

[Arezki93]

unfortunately I did not find a solution !

Le dim. 8 août 2021 à 15:29, TomerJakobovich @.***> a écrit :

I also encountered this issue, @Arezki93 https://github.com/Arezki93 did you find a solution?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/google/mediapipe/issues/2312#issuecomment-894798054, or unsubscribe https://github.com/notifications/unsubscribe-auth/AOMHEADMGTH2B4SVGH4LA53T32BDHANCNFSM5AT6KHVA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email .

[sgowroji]

Hi @Arezki93, MediaPipe tflite where updated with ops here https://github.com/google/mediapipe/blob/ecb5b5f44ab23ea620ef97a479407c699e424aa7/mediapipe/util/tflite/cpu_op_resolver.cc#:~:text=void%20MediaPipe_RegisterTfLiteOpResolver(tflite,%7D But not sure about it in python.

[jiuqiant]

MediaPipe repo has custom ops for the hair segmentation model. So, using the model with tflite python interpreter directly doesn't work. You need to copy the custom ops to a local TensorFlow repo and rebuild its python package.

Since it's not a MediaPipe Python issue, reassign this back to @sgowroji .

[sgowroji]

Closing this issue w.r.t the above comments, answers your query.

[google-ml-butler[bot]]

Are you satisfied with the resolution of your issue? Yes No