dtoolai.data¶
-
class
dtoolai.data.ImageDataSet(uri, usetype='train')[source]¶ Class allowing a collection of images annotated with categories to be used as both a Pytorch Dataset and a dtool DataSet.
-
class
dtoolai.data.TensorDataSet(uri)[source]¶ Class that allows numpy arrays to be accessed as both a pytorch Dataset and a dtool DataSet.
-
dim¶ The linear dimensions of the tensor, e.g. it is dim x dim in shape.
-
input_channels¶ The number of channels each tensor provides.
-
-
class
dtoolai.data.WrappedDataSet(uri)[source]¶ Subclass of pytorch Dataset that provides dtool DataSet methods.
This class mostly provides methods that consumers of DataSets require, and passes those methods onto its internal DataSet object.
Parameters: uri – URI for enclosed dtool DataSet.
-
dtoolai.data.coerce_to_fixed_size_rgb(im, target_dim)[source]¶ Convert a PIL image to a fixed size and 3 channel RGB format.
-
dtoolai.data.create_tensor_dataset_from_arrays(output_base_uri, output_name, data_array, label_array, image_dim, readme_content)[source]¶ Create a dtool DataSet with the necessary annotations to be used as a TensorDataSet.
Parameters: - output_base_uri – The base URI where the dataset will be created.
- output_name – The name for the output dataset.
- data_array (ndarray) – The numpy array holding data.
- label_array (ndarray) – The numpy array holding labels.
- image_dim (tuple) – Dimensions to which input images should be reshaped.
- readme_content (string) – Content that will be used to create README.yml in the created dataset.
Returns: The URI of the created dataset
Return type: URI