Marmelator/image-classification-service
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 29 Wiki Activity

beautify

Browse Source
This commit is contained in:
Julius Unverfehrt 2023-02-02 15:26:33 +01:00
parent c4416636c0
commit 94652aafe4
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
image_prediction/image_extractor/extractors/parsable.py
View File

@ -111,6 +111,7 @@ def get_metadata_for_images_on_page(page: fitz.Page):
def filter_valid_metadata(metadata):
yield from compose(
# filter_out_page_sized_images, TODO: Link to concept for extraction toggling and reclassification endpoint
filter_out_tiny_images,
filter_invalid_metadata,
)(metadata)
@ -135,8 +136,10 @@ def get_image_infos(page: fitz.Page) -> List[dict]:
@lru_cache(maxsize=None)
def xref_to_image(doc, xref) -> Image:
# NOTE: image extraction is done via pixmap to array, as this method is twice as fast as extraction via bytestream
pixmap = fitz.Pixmap(doc, xref)
array = np.frombuffer(pixmap.samples, dtype=np.uint8).reshape(pixmap.h, pixmap.w, pixmap.n)
# TODO: Find a better solution: PIL.Image.fromarray doesn't take grayscale images of the shape (h, w, 1) but (h, w)
array = array[:, :, 0] if array.shape[2] == 1 else array
return Image.fromarray(array)