Implementation of FFD using multidim_image_augmentation

[Afzal786]

Sir,

while implementing the multidim_image_augmentation repo I have came across some observations and issues:

1. Using Bazel the _augmentation_ops.so file is missing. Can you please tell me how to resolve the issue?

/content/custom-op/multidim_image_augmentation/python/ops
/content/custom-op/multidim_image_augmentation/python/ops/_augmentation_ops.so

NotFoundError Traceback (most recent call last)
in ()
----> 1 import multidim_image_augmentation

2 frames
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/load_library.py in load_op_library(library_filename)
56 RuntimeError: when unable to load the library or get the python wrappers.
57 """
---> 58 lib_handle = py_tf.TF_LoadLibrary(library_filename)
59 try:
60 wrappers = _pywrap_python_op_gen.GetPythonWrappers(

NotFoundError: /content/custom-op/multidim_image_augmentation/python/ops/_augmentation_ops.so: cannot open shared object file: No such file or directory

2. I tried to do the same using make. For that, I have created a Makefile and built the package for augmentation_ops. But while creating a cubic interpolation I am facing the issue as given below:

---

AttributeError Traceback (most recent call last)
in ()
----> 1 dense_deformation_field = augmentation_ops.cubic_interpolation2d_op(
2 control_grid, control_grid_spacings_pix, transformed_image_shape)

AttributeError: module 'ed34cdb71a1409bc10969aaf278bbfee' has no attribute 'cubic_interpolation2d_op'

Can you please tell me at least the first one how to resolve it?
It will be really helpful for me.

[Afzal786]

sir, kindly look into the above-given issues.....I will really appreciate that.

[Afzal786]

are you there sir......?

[ardila]

Hi! It seems like you need to follow the instructions to build inside of the docker container that are here. Can you give me the exact list of steps you did and I can tell you where you went wrong.

It seems like the version of the package that I made can only be pip installed with tensorflow==1.15

I expect to build a new version soon, and when I do I can post here. However for now, the only things that will work are to either downgrade to tensorflow==1.15, or to build the wheel yourself by following the instructions here.

Can you re-build the pip package and install that, but this time tell me the exact steps you followed to build the wheel?