class DataConfig(cfg.DataConfig):
"""Input config for training."""
input_path: str = ''
global_batch_size: int = 0
is_training: bool = True
dtype: str = 'float32'
shuffle_buffer_size: int = 10000
cycle_length: int = 10
is_multilabel: bool = False
aug_rand_hflip: bool = True
aug_crop: Optional[bool] = True
aug_type: Optional[
common.Augmentation] = None # Choose from AutoAugment and RandAugment.
color_jitter: float = 0.
random_erasing: Optional[common.RandomErasing] = None
file_type: str = 'tfrecord'
image_field_key: str = 'image/encoded'
label_field_key: str = 'image/class/label'
decode_jpeg_only: bool = True
mixup_and_cutmix: Optional[common.MixupAndCutmix] = None
decoder: Optional[common.DataDecoder] = common.DataDecoder()
# Keep for backward compatibility.
aug_policy: Optional[str] = None # None, 'autoaug', or 'randaug'.
randaug_magnitude: Optional[int] = 10
class DataConfig(cfg.DataConfig):
"""Input config for training."""
output_size: List[int] = dataclasses.field(default_factory=list)
# If crop_size is specified, image will be resized first to
# output_size, then crop of size crop_size will be cropped.
crop_size: List[int] = dataclasses.field(default_factory=list)
input_path: str = ''
global_batch_size: int = 0
is_training: bool = True
dtype: str = 'float32'
shuffle_buffer_size: int = 1000
cycle_length: int = 10
# If resize_eval_groundtruth is set to False, original image sizes are used
# for eval. In that case, groundtruth_padded_size has to be specified too to
# allow for batching the variable input sizes of images.
resize_eval_groundtruth: bool = True
groundtruth_padded_size: List[int] = dataclasses.field(
default_factory=list)
aug_scale_min: float = 1.0
aug_scale_max: float = 1.0
aug_rand_hflip: bool = True
preserve_aspect_ratio: bool = True
aug_policy: Optional[str] = None
drop_remainder: bool = True
file_type: str = 'tfrecord'
decoder: Optional[common.DataDecoder] = common.DataDecoder()
class DataConfig(cfg.DataConfig):
"""Input config for training."""
input_path: str = ''
global_batch_size: int = 0
is_training: bool = False
dtype: str = 'bfloat16'
decoder: common.DataDecoder = common.DataDecoder()
parser: Parser = Parser()
shuffle_buffer_size: int = 10000
file_type: str = 'tfrecord'
class DataConfig(cfg.DataConfig):
"""Input config for training."""
input_path: str = ''
tfds_name: str = ''
tfds_split: str = 'train'
global_batch_size: int = 0
is_training: bool = False
dtype: str = 'bfloat16'
decoder: common.DataDecoder = common.DataDecoder()
shuffle_buffer_size: int = 10000
file_type: str = 'tfrecord'
drop_remainder: bool = True
class DataConfig(cfg.DataConfig): """Input config for training.""" input_path: str = '' global_batch_size: int = 0 is_training: bool = False dtype: str = 'bfloat16' decoder: common.DataDecoder = common.DataDecoder() parser: Parser = Parser() shuffle_buffer_size: int = 10000 file_type: str = 'tfrecord' drop_remainder: bool = True # Number of examples in the data set, it's used to create the annotation file. num_examples: int = -1
