Handler package

class handlers.BytesHandler(delegate_to=None, shape=[])

Bases: handlers.FeatureHandler

Handler for byte features

dtype

Returns the tf data type.

This property is used when parsing a tfrecord.

handle(row, key)

Convert data dict into tf feature.

Parameters:

• row (dict) – Dictionary containing all features.
• key (str) – Name of feature to be handled.

Returns:

key to tf feature.

Return type:

dict

isfixedlen

Returns if handler works on fixed length features.

This property is used when parsing a tfrecord.

class handlers.FeatureHandler(delegate_to, shape)

Bases: object

Base class for converting data into tensorflow features

This class is used during serialization of data into tfrecords. Moreover, hanlder properties are also used when parsing tfrecords.

delegate_to

str – Name of handler that manages the feature.

shape

list – List of int containing the shape of the feature.

dtype

Returns the tf data type.

This property is used when parsing a tfrecord.

handle(row, key)

Convert data dict into tf feature.

Parameters:

• row (dict) – Dictionary containing all features.
• key (str) – Name of feature to be handled.

Returns:

key to tf feature.

Return type:

dict

isfixedlen

Returns if handler works on fixed length features.

This property is used when parsing a tfrecord.

class handlers.Float32Handler(delegate_to=None, shape=[])

Bases: handlers.FeatureHandler

Handler for float features

dtype

Returns the tf data type.

This property is used when parsing a tfrecord.

handle(row, key)

Convert data dict into tf feature.

Parameters:

• row (dict) – Dictionary containing all features.
• key (str) – Name of feature to be handled.

Returns:

key to tf feature.

Return type:

dict

isfixedlen

Returns if handler works on fixed length features.

This property is used when parsing a tfrecord.

class handlers.Int64Handler(delegate_to=None, shape=[])

Bases: handlers.FeatureHandler

Handler for integer features

dtype

Returns the tf data type.

This property is used when parsing a tfrecord.

handle(row, key)

Convert data dict into tf feature.

Parameters:

• row (dict) – Dictionary containing all features.
• key (str) – Name of feature to be handled.

Returns:

key to tf feature.

Return type:

dict

isfixedlen

Returns if handler works on fixed length features.

This property is used when parsing a tfrecord.

class handlers.NiftiHandler(img_folder, img_shape, img_dtype, img_slice=None, img_key='image')

Bases: handlers.FeatureHandler

Handler for nifti images

Loads nifti images using an img_id, potentially resizes them to img_shape, and returns a byte feature. Moreover, it adds img_shape as int feature.

img_folder

str – Path to folder containing the nifti images.

img_shape

list of int – Desired shape for the images.

img_dtype

str – Desired data type for the images.

slices

dict – Mapping of slice name to slice tuple

dtype

Returns the tf data type.

This property is used when parsing a tfrecord.

handle(row, key)

Convert data dict into tf feature.

Parameters:

• row (dict) – Dictionary containing all features.
• key (str) – Name of feature to be handled.

Returns:

key to tf feature.

Return type:

dict

isfixedlen

Returns if handler works on fixed length features.

This property is used when parsing a tfrecord.

handlers.features_from_handlers(keys_to_handlers)

Construct Fixed/VarLen Features

Used for parsing of serialized data, converts feature information (isfixedlen, shape, dtype) into TF feature.

Parameters:keys_to_handlers (dict) – Mapping of keys to handlers Returns:Mapping of keys to TF features Return type:keys_to_features (dict)