gradio-image-annotation-redaction 0.5.5

A Gradio component that can be used to annotate images with bounding boxes. Adapted for use with Document Redaction app. Forked from original work by edgarGarcia and icyray.

gradio_image_annotation_redaction

A Gradio component that can be used to annotate images with bounding boxes. Adapted for use with Document Redaction app. Forked from original work by edgarGarcia and icyray.

Installation

pip install gradio_image_annotation_redaction

Usage

import gradio as gr
from gradio_image_annotation_redaction import image_annotator
import numpy as np

example_annotation = {
    "image":   "https://raw.githubusercontent.com/seanpedrick-case/document_redaction_examples/refs/heads/main/example_complaint_letter.jpg",
    "boxes": [
        {
            "xmin": 125,
            "ymin": 239,
            "xmax": 230,
            "ymax": 266,
            "label": "Name",
            "color": (255, 0, 0),
            "text": "Mark Smith",
            "page": 1,
        },
        {
            "xmin": 125,
            "ymin": 288,
            "xmax": 301,
            "ymax": 375,
            "label": "Address",
            "color": (0, 255, 0),
            "text": "Sofa Showroom, 555 Broadway, Cityland, KS 66214",
            "page": 1,
        }
    ]
}

examples_crop = [
    {
        "image": "https://raw.githubusercontent.com/gradio-app/gradio/main/guides/assets/logo.png",
        "boxes": [
            {
                "xmin": 30,
                "ymin": 70,
                "xmax": 530,
                "ymax": 500,
                "color": (100, 200, 255),
            }
        ],
    },
    {
        "image": "https://raw.githubusercontent.com/seanpedrick-case/document_redaction_examples/refs/heads/main/example_complaint_letter.jpg",
        "boxes": [
            {
                "xmin": 636,
                "ymin": 575,
                "xmax": 801,
                "ymax": 697,
                "color": (255, 0, 0),
            },
        ],
    },
]


def crop(annotations:dict):
    if angle:= annotations.get("orientation", None):
        annotations["image"] = np.rot90(annotations["image"], k=-angle)
    if annotations["boxes"]:
        box = annotations["boxes"][0]
        return annotations["image"][
            box["ymin"]:box["ymax"],
            box["xmin"]:box["xmax"]
        ]
    return None

def _image_size(image):
    """Derive (width, height) from image when numpy or PIL."""
    if image is None:
        return None, None
    if isinstance(image, np.ndarray):
        if image.ndim >= 2:
            return int(image.shape[1]), int(image.shape[0])  # width, height
        return None, None
    if hasattr(image, "size"):  # PIL Image
        w, h = image.size
        return int(w), int(h)
    return None, None


def get_boxes_json(annotations):
    if annotations is None:
        return None
    image = annotations.get("image")
    image_path = image if isinstance(image, str) else None
    # Use frontend-provided dimensions when present, else derive from image data (numpy/PIL)
    image_width = annotations.get("image_width")
    image_height = annotations.get("image_height")
    if image_width is None or image_height is None:
        w, h = _image_size(image)
        if image_width is None:
            image_width = w
        if image_height is None:
            image_height = h
    return {
        "boxes": annotations.get("boxes", []),
        "orientation": annotations.get("orientation"),
        "image_width": image_width,
        "image_height": image_height,
        "image_path": image_path,
    }


with gr.Blocks() as demo:
    with gr.Tab("Object annotation", id="tab_object_annotation"):
        
        annotator = image_annotator(
            example_annotation,
            label_list=["Person", "Vehicle"],
            label_colors=[(0, 255, 0), (255, 0, 0)],
            image_type="filepath",
        )
        button_get = gr.Button("Get bounding boxes")
        json_boxes = gr.JSON()
        button_get.click(get_boxes_json, annotator, json_boxes)

    with gr.Tab("Crop", id="tab_crop"):
        with gr.Row():
            annotator_crop = image_annotator(
                examples_crop[0],
                image_type="numpy",
                disable_edit_boxes=True,
                single_box=True,
            )
            image_crop = gr.Image()
        button_crop = gr.Button("Crop")
        button_crop.click(crop, annotator_crop, image_crop)

        gr.Examples(examples_crop, annotator_crop)
    
    with gr.Accordion("Keyboard Shortcuts"):
        gr.Markdown("""
        - ``C``: Create mode
        - ``D``: Drag mode
        - ``E``: Edit selected box (same as double-click a box)
        - ``Delete``: Remove selected box
        - ``Space``: Reset view (zoom/pan)
        - ``Enter``: Confirm modal dialog
        - ``Escape``: Cancel/close modal dialog
        """)

if __name__ == "__main__":
    demo.launch()

image_annotator

Initialization

name	type	default	description
value	dict| None	value = None	A dict or None. The dictionary must contain a key 'image' with either an URL to an image, a numpy image or a PIL image. Optionally it may contain a key 'boxes' with a list of boxes. Each box must be a dict wit the keys: 'xmin', 'ymin', 'xmax' and 'ymax' with the absolute image coordinates of the box. Optionally can also include the keys 'label' and 'color' describing the label and color of the box. Color must be a tuple of RGB values (e.g. `(255,255,255)`). Optionally can also include the keys 'orientation' with a integer between 0 and 3, describing the number of times the image is rotated by 90 degrees in frontend, the rotation is clockwise.
boxes_alpha	float| None	value = None	Opacity of the bounding boxes 0 and 1.
label_list	list[str]| None	value = None	List of valid labels.
label_colors	list[str]| None	value = None	Optional list of colors for each label when `label_list` is used. Colors must be a tuple of RGB values (e.g. `(255,255,255)`).
box_min_size	int| None	value = None	Minimum valid bounding box size.
handle_size	int| None	value = None	Size of the bounding box resize handles.
box_thickness	int| None	value = None	Thickness of the bounding box outline.
box_selected_thickness	int| None	value = None	Thickness of the bounding box outline when it is selected.
disable_edit_boxes	bool| None	value = None	Disables the ability to set and edit the label and color of the boxes.
single_box	bool	value = False	If True, at most one box can be drawn.
height	int| str| None	value = None	The height of the displayed image, specified in pixels if a number is passed, or in CSS units if a string is passed.
width	int| str| None	value = None	The width of the displayed image, specified in pixels if a number is passed, or in CSS units if a string is passed.
image_mode	"1"| "L"| "P"| "RGB"| "RGBA"| "CMYK"| "YCbCr"| "LAB"| "HSV"| "I"| "F"	value = "RGB"	"RGB" if color, or "L" if black and white. See https://pillow.readthedocs.io/en/stable/handbook/concepts.html for other supported image modes and their meaning.
sources	list["upload"| "webcam"| "clipboard"]| None	value = ['upload', 'webcam', 'clipboard']	List of sources for the image. "upload" creates a box where user can drop an image file, "webcam" allows user to take snapshot from their webcam, "clipboard" allows users to paste an image from the clipboard. If None, defaults to ["upload", "webcam", "clipboard"].
image_type	"numpy"| "pil"| "filepath"	value = "numpy"	The format the image is converted before being passed into the prediction function. "numpy" converts the image to a numpy array with shape (height, width, 3) and values from 0 to 255, "pil" converts the image to a PIL image object, "filepath" passes a str path to a temporary file containing the image. If the image is SVG, the `type` is ignored and the filepath of the SVG is returned.
label	str| None	value = None	The label for this component. Appears above the component and is also used as the header if there are a table of examples for this component. If None and used in a `gr.Interface`, the label will be the name of the parameter this component is assigned to.
container	bool	value = True	If True, will place the component in a container - providing some extra padding around the border.
scale	int| None	value = None	relative size compared to adjacent Components. For example if Components A and B are in a Row, and A has scale=2, and B has scale=1, A will be twice as wide as B. Should be an integer. scale applies in Rows, and to top-level Components in Blocks where fill_height=True.
min_width	int	value = 160	minimum pixel width, will wrap if not sufficient screen space to satisfy this value. If a certain scale value results in this Component being narrower than min_width, the min_width parameter will be respected first.
interactive	bool| None	value = True	if True, will allow users to upload and annotate an image; if False, can only be used to display annotated images.
visible	bool	value = True	If False, component will be hidden.
elem_id	str| None	value = None	An optional string that is assigned as the id of this component in the HTML DOM. Can be used for targeting CSS styles.
elem_classes	list[str]| str| None	value = None	An optional list of strings that are assigned as the classes of this component in the HTML DOM. Can be used for targeting CSS styles.
render	bool	value = True	If False, component will not render be rendered in the Blocks context. Should be used if the intention is to assign event listeners now but render the component later.
show_label	bool| None	value = None	if True, will display label.
show_download_button	bool	value = True	If True, will show a button to download the image.
show_share_button	bool| None	value = None	If True, will show a share icon in the corner of the component that allows user to share outputs to Hugging Face Spaces Discussions. If False, icon does not appear. If set to None (default behavior), then the icon appears if this Gradio app is launched on Spaces, but not otherwise.
show_clear_button	bool| None	value = True	If True, will show a button to clear the current image.
show_remove_button	bool| None	value = None	If True, will show a button to remove the selected bounding box.
handles_cursor	bool| None	value = True	If True, the cursor will change when hovering over box handles in drag mode. Can be CPU-intensive.
use_default_label	bool	value = False	If True, the first item in label_list will be used as the default label when creating boxes.
enable_keyboard_shortcuts	bool	value = True	If True, the component will respond to keyboard events.

Events

name	description
clear	This listener is triggered when the user clears the image_annotator using the clear button for the component.
change	Triggered when the value of the image_annotator changes either because of user input (e.g. a user types in a textbox) OR because of a function update (e.g. an image receives a value from the output of an event trigger). See .input() for a listener that is only triggered by user input.
upload	This listener is triggered when the user uploads a file into the image_annotator.

User function

The impact on the users predict function varies depending on whether the component is used as an input or output for an event (or both).

• When used as an Input, the component only impacts the input signature of the user function.
• When used as an output, the component only impacts the return signature of the user function.

The code snippet below is accurate in cases where the component is used as both an input and an output.

• As output: Is passed, a dict with the image and boxes or None.
• As input: Should return, a dict with an image and an optional list of boxes or None.

def predict(
    value: AnnotatedImageValue| None
) -> AnnotatedImageValue| None:
    return value

AnnotatedImageValue

class AnnotatedImageValue(TypedDict):
    image: Optional[np.ndarray | PIL.Image.Image | str]
    boxes: Optional[List[dict]]
    orientation: Optional[int]
    image_width: Optional[int]
    image_height: Optional[int]