DeprecationWarning thrown in utils.image_utils

[johnthagen]

System information.

• Have I written custom code (as opposed to using a stock example script provided in Keras): Yes, simple.
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): macOS 11.6
• TensorFlow installed from (source or binary): binary
• TensorFlow version (use command below): v2.9.0-18-gd8ce9f9c301 2.9.1
• Python version: 3.19.13
• Bazel version (if compiling from source): N/A
• GPU model and memory: AMD Radeon Pro 5500M 4 GB
• Exact command to reproduce:

import tensorflow.lite as tflite

python -Werror main.py

Describe the problem.

Teras is using a deprecated feature of Pillow.

https://github.com/keras-team/keras/blob/3613c3defc39c236fb1592c4f7ba1a9cc887343a/keras/utils/image_utils.py#L38

Describe the current behavior.

    import tensorflow.lite as tflite
python3.9/site-packages/tensorflow/__init__.py:473: in <module>
    keras._load()
python3.9/site-packages/tensorflow/python/util/lazy_loader.py:41: in _load
    module = importlib.import_module(self.__name__)
/usr/local/Cellar/[email protected]/3.9.13_1/Frameworks/Python.framework/Versions/3.9/lib/python3.9/importlib/__init__.py:127: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
python3.9/site-packages/keras/__init__.py:24: in <module>
    from keras import models
python3.9/site-packages/keras/models/__init__.py:18: in <module>
    from keras.engine.functional import Functional
python3.9/site-packages/keras/engine/functional.py:23: in <module>
    from keras import backend
python3.9/site-packages/keras/backend.py:43: in <module>
    from keras.engine import keras_tensor
python3.9/site-packages/keras/engine/keras_tensor.py:17: in <module>
    from keras.utils import object_identity
python3.9/site-packages/keras/utils/__init__.py:20: in <module>
    from keras.utils.image_dataset import image_dataset_from_directory
python3.9/site-packages/keras/utils/image_dataset.py:22: in <module>
    from keras.utils import image_utils
python3.9/site-packages/keras/utils/image_utils.py:36: in <module>
    'nearest': pil_image.NEAREST,
python3.9/site-packages/PIL/Image.py:67: in __getattr__
    warnings.warn(
E   DeprecationWarning: NEAREST is deprecated and will be removed in Pillow 10 (2023-07-01). Use Resampling.NEAREST or Dither.NONE instead.

I am testing with Pillow 9.1.1.

Describe the expected behavior.

No DeprecationWarnings are raised when using keras.

Contributing.

• Do you want to contribute a PR? (yes/no):
• If yes, please read this page for instructions
• Briefly describe your candidate solution(if contributing):

[gadagashwini]

Hi @johnthagen, Could you share the minimal reproducible code snippet to investigate the reported issue. Thank you!

[johnthagen]

@gadagashwini Please see the code and command under "Exact command to reproduce:" in the original post.

[gadagashwini]

Hi @johnthagen, I executed import tensorflow.lite as tflite on colab I didn’t get any error. Request you to try again. Thank you!

[johnthagen]

@gadagashwini Did you execute the Python script with -Werror?

python -Werror main.py

[gadagashwini]

Hi @johnthagen, Could you share the main.py file. Thank you

[johnthagen]

1. pip install tensorflow into a Python virtual environment
2. Create a file named main.py
3. Within this file, add the contents:

import tensorflow

4. Execute the script with -Werror

python -Werror main.py

This will show the DeprecationWarning noted in the original comment. I also linked to the line of code in this repo where the deprecated usage is in the original comment.

My exact environment (tested on macOS x86_64):

Package                      Version
---------------------------- -----------
absl-py                      1.1.0
astunparse                   1.6.3
cachetools                   5.2.0
certifi                      2022.5.18.1
charset-normalizer           2.0.12
flatbuffers                  1.12
gast                         0.4.0
google-auth                  2.7.0
google-auth-oauthlib         0.4.6
google-pasta                 0.2.0
grpcio                       1.46.3
h5py                         3.7.0
idna                         3.3
importlib-metadata           4.11.4
keras                        2.9.0
Keras-Preprocessing          1.1.2
libclang                     14.0.1
Markdown                     3.3.7
numpy                        1.22.4
oauthlib                     3.2.0
opt-einsum                   3.3.0
packaging                    21.3
pip                          22.1.2
protobuf                     3.19.4
pyasn1                       0.4.8
pyasn1-modules               0.2.8
pyparsing                    3.0.9
requests                     2.28.0
requests-oauthlib            1.3.1
rsa                          4.8
setuptools                   62.3.3
six                          1.16.0
tensorboard                  2.9.1
tensorboard-data-server      0.6.1
tensorboard-plugin-wit       1.8.1
tensorflow                   2.9.1
tensorflow-estimator         2.9.0
tensorflow-io-gcs-filesystem 0.26.0
termcolor                    1.1.0
typing_extensions            4.2.0
urllib3                      1.26.9
Werkzeug                     2.1.2
wheel                        0.37.1
wrapt                        1.14.1
zipp                         3.8.0

[gadagashwini]

Hi @johnthagen, You can safely ignore these deprecated warnings.
Suppress these warnings by turning off

import logging, os
logging.disable(logging.WARNING)
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
import tensorflow as tf

Thank you!

[johnthagen]

@gadagashwini Yes, we are suppressing them, but since this will eventually cause breakage with a future Pillow release, it seemed like something that should be addressed in keras?

[leonardschneider]

This shows up in pytest warnings as well :(

[haifeng-jin]

It is fixed by keras-team/keras#16746 on master.

[google-ml-butler]

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