minerva.data.readers.multi_reader
Classes
Reader that composes items from other readers. |
Module Contents
- class minerva.data.readers.multi_reader.MultiReader(readers, preprocess=None, collate_fn=np.stack)[source]
Bases:
minerva.data.readers._ReaderReader that composes items from other readers.
Its i-th item is the i-th item of each of the child-readers merged together according to a collate_fn function.
Collects data from multiple readers and collates them
Parameters
- readers: Sequence[_Reader]
The readers from which the data will be collected. At least one must be provided. If the readers have different lengths, data will only be collected up until the length of the smallest child-reader.
- preprocess: Optional[Callable]
A function to be applied individually to each item read from the child-readers. Defaults to an identity function (i.e. no changes to the data).
- collate_fn: Optional[Callable]
A function that recieves a list of items read from the child-readers after preprocessing and returns a single item for this reader. Defaults to numpy.stack, which means it must be provided if the preprocessing function does not always return same-shape numpy arrays.
- __getitem__(index)[source]
Retrieves the items from each reader at the specified index and collates them accordingly.
Parameters
- indexint
Index of the item to retrieve.
Returns
- Any
An item from the reader.
- Parameters:
index (int)
- Return type:
Any
- __len__()[source]
Returns the length the reader, defined as the length of the smallest child-reader
Returns
- int
The length of the reader.
- Return type:
int
- _readers
- collate_fn
- preprocess
- Parameters:
readers (Sequence[minerva.data.readers._Reader])
preprocess (Optional[Callable])
collate_fn (Optional[Callable])