gmft.table_visualization module

gmft.table_visualization.display_html_and_image(html_content, pil_image)

This is strictly for Jupyter notebooks. It displays the HTML content and the PIL image side by side.

gmft.table_visualization.plot_interval_histogram(histogram, figsize=(12, 6), invert_x=False, dotted_line_at=1)

Plot an interval histogram showing frequency changes and optionally the original intervals.

Args:

histogram: IntervalHistogram instance to plot show_intervals: If True, show the original intervals below the histogram figsize: Tuple of (width, height) for the figure

Returns:

matplotlib figure object

gmft.table_visualization.plot_results_orig(pil_img, results, id2label, filter=None)

Takes tensor input. results = {

“scores”: tensor([0.993, 0.927]), “labels”: tensor([0, 0]), “boxes”: tensor([[0.000, 0.000, 70.333, 20.333], # bounding boxes: xmin, ymin, xmax, ymax

[10.001, 0.001, 0.998, 0.998]]),

}

gmft.table_visualization.plot_results_unwr(pil_img, confidence: List[float], labels: List[int], boxes: List[Tuple[float, float, float, float]], id2label: dict[int, str], filter: List[int] = None, figsize: Tuple[int, int] = (32, 20), padding: Tuple[int, int] | None = None, margin: Tuple[int, int] | None = None, linewidth: int = 3, show_labels: bool = True, return_img: bool = False)

Helper method to visualize the results of the table detection/format model.

Parameters:

  • pil_img – PIL image

  • confidence – list of floats, confidence scores

  • labels – list of integers, class labels

  • boxes – list of lists, bounding boxes in the format [xmin, ymin, xmax, ymax]

  • id2label – dictionary, mapping class labels (int) to class names

  • filter – list of integers, class labels to selectively display

  • figsize – tuple, figure size. None for a smaller size

  • show_labels – boolean, whether to display the class labels

Example:

confidence = [0.993, 0.927]
labels = [0, 0]  # 0 is the table class
boxes = [
    [0.000, 0.000, 70.333, 20.333],  # bounding boxes: xmin, ymin, xmax, ymax
    [10.001, 0.001, 0.998, 0.998]
]
gmft.table_visualization.plot_shaded_boxes(pil_img, labels: list[int], boxes: list[tuple[float, float, float, float]], id2color: dict = None, filter=None, alpha=0.2, id2border: dict = None)

Helper method to visualize the results of the table detection/format model using PIL.

Parameters:

  • pil_img – PIL image

  • labels – list of integers, class labels

  • boxes – list of tuples, bounding boxes in the format [xmin, ymin, xmax, ymax]

  • id2color – dictionary, mapping class labels (int) to desired color

  • filter – list of integers, class labels to selectively display

  • alpha – float, transparency for the box fill (0 to 1)