Problem with image_dataset_from_directory

General Discussion
,
#1

I’m using Kaggle with TensorFlow 2.12. This code:

train_ds, val_ds = tf.keras.utils.image_dataset_from_directory(...)

give me this error:

---------------------------------------------------------------------------
InvalidArgumentError                      Traceback (most recent call last)
Cell In[82], line 77
     75 # Addestrare il modello
     76 tf.random.set_seed(666)
---> 77 history = model.fit(
     78     train_ds,
     79     epochs = 300,
     80     batch_size = 32,
     81     validation_data = val_ds,
     82     callbacks = [checkpoint, cos_decay] #  reduce_lr_callback
     83 )
     84 model.evaluate(val_generator) # 69

File /opt/conda/lib/python3.10/site-packages/keras/src/utils/traceback_utils.py:70, in filter_traceback.<locals>.error_handler(*args, **kwargs)
     67     filtered_tb = _process_traceback_frames(e.__traceback__)
     68     # To get the full stack trace, call:
     69     # `tf.debugging.disable_traceback_filtering()`
---> 70     raise e.with_traceback(filtered_tb) from None
     71 finally:
     72     del filtered_tb

File /opt/conda/lib/python3.10/site-packages/tensorflow/python/eager/execute.py:53, in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
     51 try:
     52   ctx.ensure_initialized()
---> 53   tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
     54                                       inputs, attrs, num_outputs)
     55 except core._NotOkStatusException as e:
     56   if name is not None:

InvalidArgumentError: Graph execution error:

Unknown image file format. One of JPEG, PNG, GIF, BMP required.
	 [[{{node decode_image/DecodeImage}}]]
	 [[IteratorGetNext]] [Op:__inference_train_function_334279]

But with ImageDataGenerator.flow_from_dataframe don’t give me any error. How to setup tf.keras.utils.image_dataset_from_directory to avoid this problem.

#3

@Automata,

Welcome to the Tensorflow Forum,

Even with the correct extension, some images can still be of poor quality due to the presence of bad binaries resulting in a corrupted image. You can try to filter out corrupted images as shown here.

Thank you!

1 Like
#5

Thanks.
But why doesn’t ImageDataGenerator.flow_from_dataframe give me this problem?

#6

Yes, it should work. But without more details, it’s not clear what might be happening here.

Thank you!

#8

This did not work for me.

Is there any other solutions?