Computes the aggregated metrics Returns: a tuple with the recall & precision for each string comparison and the mean IoU

Args: in_channels: number of input channels out_dim: final embedding dimension embed_dims: list of embedd

Remove padding from the localization predictions Args: pages: list of pages loc_preds: list of localization predictions p

Order boxes to group them in lines Args: boxes: bounding boxes of shape (N, 4) or (N, 4, 2) in case of rotated bbox Retu

Split a line in sub_lines Args: boxes: bounding boxes of shape (N, 4) word_idcs: list of indexes for the words of the

Save model and config to disk for pushing to huggingface hub Args: model: PyTorch model to be saved save_dir: directory to save m

Internal constructor for the VIPNet models. Args: arch: architecture key pretrained: load pretrained weights? ignore

Add a hook to the predictor Args: hook: a callable that takes as input the `loc_preds` and returns the modified `loc_preds`

Encode a list of gts sequences into a np array and gives the corresponding* sequence lengths. Args: gts: list of ground-t

Compute the shape of an expanded rotated image Args: img_shape: the height and width of the image angle: angle between -90 and +9

Generate predictions for the given features.

Given a predefined mapping, decode the sequence of numbers to a string Args: input_seq: array to decode mapping: vocabulary (stri

Detach the objectness scores from box predictions Args: boxes: list of arrays with boxes of shape (N, 5) or (N, 5, 2) Returns:

Created cropped images from list of bounding boxes Args: img: input image boxes: bounding boxes of shape (N, 4) where N is the nu

Created cropped images from list of rotated bounding boxes Args: img: input image polys: bounding boxes of shape (N, 4, 2)

Re-arrange features into cross-shaped windows for Q/K. Args: x: A float tensor of shape (b, n, c). size: A t

End-to-end KIE architecture using one model for localization, and another for text recognition. >>> import numpy as np >>> from doctr.models

Merges consecutive string sequences with overlapping characters. Args: seq_list: list of sequences to merge. Sequences need to be or

Merges 2 character sequences in the best way to maximize the alignment of their overlapping characters. Args: a: first char seq, suffix s

Display the results of the LR grid search. Adapted from https://github.com/frgfm/Holocron/blob/master/holocron/trainer/core.py Args:

Create a matplotlib polygon patch for the element Args: geometry: bounding box of the element page_dimensions: dimensions of the

Expand a polygon (points) by a factor unclip_ratio, and returns a polygon Args: points: The first parameter. Returns:

Expand a polygon (points) by a factor unclip_ratio, and returns a polygon Args: points: The first parameter. Returns:

Postprocess the model output Args: output: the model output to postprocess input_images: the input images us

Preprocess the input image Args: img: the input image to preprocess Returns: np.ndarray: the prepro

Apply a random shadow effect to an image using NumPy for blurring. Args: img: Image to modify (C, H, W) as a PyTorch tensor. opac

Reconstruct predictions from possibly split crops. Args: preds: List of (text, confidence) tuples from each crop. crop_map:

Rotate a batch of bounding boxes or polygons by an angle around the image center. Args: geoms: (N, 4) or (N, 4, 2) array of ABSOLUTE

Rotate points counter-clockwise. Args: points: array of size (N, 2) angle: angle between -90 and +90 degrees Returns:

Rotate an image counterclockwise by an given angle. Args: image: numpy tensor to rotate angle: rotation angle in degrees, between

Performs string comparison with multiple levels of tolerance Args: word1: a string word2: another string Returns: a

Updates the metric Args: gt_boxes: a set of relative bounding boxes either of shape (N, 4) or (N, 5) if they are rotated ones

Visualize a full page with predicted blocks, lines and words >>> import numpy as np >>> import matplotlib.pyplot as plt >>> from doctr.ut

Gather independent words in structured blocks Args: boxes: bounding boxes of all detected words of the page, of shape (N, 4) or (

Gather independent words in structured blocks Args: boxes: bounding boxes of all detected words of the page, of shape (N, 4) or (

Download the model from the given url if needed Args: url: the url to use model_path: the path to the model

Order lines to group them in blocks Args: boxes: bounding boxes of shape (N, 4) or (N, 4, 2) lines: list of lines, ea