没有注册任何OpKernel以在iOS上使用这些attrs支持Op'DecodeJpeg'

Question

I have followed the steps from Tensorflow For Poets ( link ) to train my own dataset using tensorflow, and getting very good prediction result.This provides me retrained.py, label_images.py, retrained_graph.pb, retrained_labels.txt files.So i have my own model graph and label file.

After that i have downloaded Tensorflow Example Project for iOS from github. The project runs successfully in iOS platform. but when i am using my own retrained_graph.pb, retrained_labels.txt file on behalf of examples model and label file , i am getting error as following -

Errors :

Could not create TensorFlow Graph: Invalid argument: No OpKernel was registered to support Op 'DecodeJpeg' with these attrs. Registered devices: [CPU], Registered kernels:

[[Node: DecodeJpeg = DecodeJpegacceptable_fraction=1, channels=3, dct_method="", fancy_upscaling=true, ratio=1, try_recover_truncated=false]]

and ,

Couldn't load model: Invalid argument: No OpKernel was registered to support Op 'DecodeJpeg' with these attrs. Registered devices: [CPU], Registered kernels:

[[Node: DecodeJpeg = DecodeJpegacceptable_fraction=1, channels=3, dct_method="", fancy_upscaling=true, ratio=1, try_recover_truncated=false]]

Can you help me please ? I have been searching a long time for it . thanks

Answer 1

You're placing the image decoding operation on the GPU, while this op is available only for the execution on the CPU.

From the documentation

If you would like TensorFlow to automatically choose an existing and supported device to run the operations in case the specified one doesn't exist, you can set allow_soft_placement to True in the configuration option when creating the session.

therefore, you simply have to change your session initialization with

 tf.Session(config=tf.ConfigProto(
  allow_soft_placement=True))

Answer 2

TensorFlow on iOS is a Subset of PC's, It doesn't contain some functions that use too much memory.

You can transform you pb-file to which CoreML supports.

This Tool may help you tf-coreml