def __init__(self,
data_path,
config=None,
datatype="categorical",
training=True):
"""__init__.
Args:
data_path: Dataset Path which should be in structure way
please see readme file for more details on structuring.
config: Config file , a dict file contains all required attributes
to configure.
datatype: Dataset Type i.e (Bbox , Labels ,Segmentations)
bbox dataset is object detection dataset which will be provided in
form of [image,bboxs] or [image, class_targets,bbox_targets].
training: is pipeline in training mode or not?
"""
# bunch the config dict.
config = Bunch(config)
if not isinstance(data_path, str):
msg = f"datapath should be str but pass {type(data_path)}."
logging.error(msg)
raise TypeError("Only str allowed")
self._datatype = datatype
self._data_path = data_path
self.config = config
self._training = training
self._shuffle_buffer = None
self._batch_size = self.config.get("batch_size", 32)
self._image_size = self.config.get("image_size", [512, 512])
self._drop_remainder = self.config.get("drop_remainder", True)
self.augmenter = augment.Augment(self.config, datatype)
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
# Create an Augmentation pipeline !
config = {
"batch_size": 1,
"image_size": [512, 512],
"transformations": {
"flip_left_right": None,
"gridmask": None,
"random_rotate": None,
},
"categorical_encoding": "labelencoder",
}
config = Bunch(config)
self.augmentor = augment.Augment(config)
tf.compat.v1.random.set_random_seed(111111)