page.synthesize() returns error when assume_straight_pages=False

[hienphan161]

Bug description

I've seen that when I use assume_straight_pages=False, it makes result.pages[0].synthesize() stop working.

Code snippet to reproduce the bug

predictor = ocr_predictor(
  pretrained=True,
  assume_straight_pages=False,
  straighten_pages=True,
  detect_orientation=True,
  resolve_blocks=True
)

file_name = "/content/vecteezy_modern-professional-business-card-creative-and-simple_9010481 - Copy.jpg"
image = Image.open(file_name)
image = ImageOps.exif_transpose(image)

result = predictor([np.asarray(image)])
result.pages[0].synthesize()

Error traceback

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
[<ipython-input-7-7d5fc2bce17c>](https://localhost:8080/#) in <cell line: 1>()
----> 1 result.pages[0].synthesize()

1 frames
[/usr/local/lib/python3.10/dist-packages/doctr/io/elements.py](https://localhost:8080/#) in synthesize(self, **kwargs)
    315             synthesized page
    316         """
--> 317         return synthesize_page(self.export(), **kwargs)
    318 
    319     def export_as_xml(self, file_title: str = "docTR - XML export (hOCR)") -> Tuple[bytes, ET.ElementTree]:

[/usr/local/lib/python3.10/dist-packages/doctr/utils/reconstitution.py](https://localhost:8080/#) in synthesize_page(page, draw_proba, font_family)
     41             for word in line["words"]:
     42                 # Get absolute word geometry
---> 43                 (xmin, ymin), (xmax, ymax) = word["geometry"]
     44                 xmin, xmax = int(round(w * xmin)), int(round(w * xmax))
     45                 ymin, ymax = int(round(h * ymin)), int(round(h * ymax))

ValueError: too many values to unpack (expected 2)

Environment

DocTR version: v0.9.0
TensorFlow version: 2.17.0
PyTorch version: 2.3.1+cu121 (torchvision 0.18.1+cu121)
OpenCV version: 4.10.0
OS: Ubuntu 22.04.3 LTS
Python version: 3.10.12
Is CUDA available (TensorFlow): Yes
Is CUDA available (PyTorch): Yes
CUDA runtime version: 12.2.140
GPU models and configuration: GPU 0: Tesla T4
Nvidia driver version: 535.104.05
cuDNN version: Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.8.9.6
/usr/lib/x86_64-linux-gnu/libcudnn_adv_infer.so.8.9.6
/usr/lib/x86_64-linux-gnu/libcudnn_adv_train.so.8.9.6
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_infer.so.8.9.6
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_train.so.8.9.6
/usr/lib/x86_64-linux-gnu/libcudnn_ops_infer.so.8.9.6
/usr/lib/x86_64-linux-gnu/libcudnn_ops_train.so.8.9.6

Deep Learning backend

is_tf_available: False
is_torch_available: True

[felixdittrich92]

Hi @hienphan161 👋,

Thanks for reporting.

You are right.
We don't have logic to reconstitute polygons (possible rotated texts) in synthesize() yet.
So there is a missing check for the geometry or exception handling. 👍

[felixdittrich92]

#1750