def test_extract():
# check if it works through both types
post = rp.extract(test_img)
post = rp.extract("tests/test_resources/test_receipt.jpeg")
# check type
assert isinstance(post, tuple)
def test_extract_lines():
# test extraction from folder of images
from_folder = rp.extract("tests/test_resources/test_receipt.jpeg")
assert isinstance(from_folder, tuple)
assert isinstance(from_folder[0], type(np.array([])))
assert len(from_folder[0].shape) == 3
img = cv2.imread("tests/test_resources/test_receipt.jpeg")
from_cv2 = rp.extract(img)
assert isinstance(from_cv2, tuple)
assert isinstance(from_cv2[0], type(np.array([])))
assert len(from_cv2[0].shape) == 3
def test_ocr_item_pipe():
imgs = rp.extract("tests/test_resources/test_receipt.jpeg")
items = rp.item_pipe(imgs)
# check type
assert isinstance(items, type(pd.DataFrame([1, 2, 3])))
assert len(imgs) == len(items)
assert len(items.columns) == 2
def test_ocr_text_extract():
imgs = rp.extract("tests/test_resources/test_receipt.jpeg")
result = rp.ocr.ocr_textM(imgs)
assert len(imgs) == len(result)
assert isinstance(result[0][0], str)
assert isinstance(result[0][1], type(np.zeros(1)))
assert all([isinstance(e, str) for e in rp.text._text(result)])
# sk_model & model_params can be passed see documentation ) # -------------------------------------------------------------------------------- # Example of extracting lines from an image # If no classifier path is provided a default classifier will be used # which is only trained on one store type and has loose ground truth # labels (see docs on how to provide a classifier from previous step) # NOTE: As with the ``make_training_data`` function above, you can and should # use the context dictionary for extraction. It would ideally be the same as the one # used to train the model for a given shop. So the padding and other parameters # are respected. # From a image path: imgs = rp.extract(input_image_as_str) # From Image imgs = rp.extract(input_img_loaded_with_io_load_validate) # From a path to a folder imgs = rp.extract_from_folder(folder_path) # I've added aditional functionality! # If you run the following command, you will get a tuple out (items, non-items). imgs, non_items = rp.extract(input_see_above, return_negatives=True) # Afterwards you can proceed to run the rest of the pipeline as normal on the 'imgs' # However please see further down how to use the `non_items` to get date and shop name # Please note! That if you specify `override_prediction = True` in the extract # function, the line classification will be avoided and one can see just what the # text line detection is finding in terms of lines.
# third party
import cv2
# project
from readpyne import extract, ocr, io
lines = extract("tests/test_resources/test_receipt.jpeg")
def test_ocr():
assert True