List argument 'input_datasets' to 'ZipDataset' Op with length 0 shorter than minimum length 1.

[yhetman]

Hello! First, I got an error with anchors in buil functio in model.py. I fixed that with the help of Anchors Layer as was adviced in issue #1930 .

And now I faced with another error :

/usr/local/lib/python3.7/dist-packages/mask_rcnn-2.1-py3.7.egg/mrcnn/model.py:659 call  *
        outputs = utils.batch_slice(
    /usr/local/lib/python3.7/dist-packages/mask_rcnn-2.1-py3.7.egg/mrcnn/utils.py:827 batch_slice  *
        outputs = list(zip(*outputs))
    /usr/local/lib/python3.7/dist-packages/tensorflow/python/autograph/operators/py_builtins.py:374 zip_  **
        return _tf_dataset_zip(*iterables)
    /usr/local/lib/python3.7/dist-packages/tensorflow/python/autograph/operators/py_builtins.py:379 _tf_dataset_zip
        return dataset_ops.DatasetV2.zip(iterables)
    /usr/local/lib/python3.7/dist-packages/tensorflow/python/data/ops/dataset_ops.py:958 zip
        return ZipDataset(datasets)
    /usr/local/lib/python3.7/dist-packages/tensorflow/python/data/ops/dataset_ops.py:3341 __init__
        **self._flat_structure)
    /usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/gen_dataset_ops.py:6578 zip_dataset
        name=name)
    /usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/op_def_library.py:530 _apply_op_helper
        (input_name, op_type_name, len(values), num_attr.minimum))

    ValueError: List argument 'input_datasets' to 'ZipDataset' Op with length 0 shorter than minimum length 1.

My code :

class TagConfig(Config):
    NAME = "tag_cng"
    NUM_CLASSES = 2
    STEPS_PER_EPOCH = 131
config = TagConfig()
model = MaskRCNN(mode='training', model_dir= './' , config=config)

Thank you!

[innovativeC]

Hi @yhetman, I am facing the same issue while initiating the MaskRCNN class. Were you able to figure out about how to make it work ?

[yhetman]

@innovativeC

Hello! Unfortunately, I haven't figured it out. Instead, I decided to use model that is provided in TensorflowAPI. Hope, you're more lucky in this issue :)

[FA1-cyber]

@innovativeC did you find a solution to this problem?