File size: 15,582 Bytes
8fdc036
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
"""gr.AnnotatedImage() component."""

from __future__ import annotations

from typing import Any, List

import gradio_client.utils as client_utils
import numpy as np
import PIL.Image
from gradio_client import file
from gradio_client.documentation import document

from gradio import processing_utils, utils
from gradio.components.base import Component
from gradio.data_classes import FileData, GradioModel
from gradio.events import Events

PIL.Image.init()  # fixes https://github.com/gradio-app/gradio/issues/2843


class Annotation(GradioModel):
    image: FileData
    label: str


class AnnotatedImageData(GradioModel):
    image: FileData
    annotations: List[Annotation]

from gradio.events import Dependency

@document()
class AnnotatedImage(Component):
    """
    Creates a component to displays a base image and colored annotations on top of that image. Annotations can take the from of rectangles (e.g. object detection) or masks (e.g. image segmentation).
    As this component does not accept user input, it is rarely used as an input component.

    Demos: image_segmentation
    """

    EVENTS = [Events.select]

    data_model = AnnotatedImageData

    def __init__(
        self,
        value: (
            tuple[
                np.ndarray | PIL.Image.Image | str,
                list[tuple[np.ndarray | tuple[int, int, int, int], str]],
            ]
            | None
        ) = None,
        *,
        format: str = "webp",
        show_legend: bool = True,
        height: int | str | None = None,
        width: int | str | None = None,
        color_map: dict[str, str] | None = None,
        label: str | None = None,
        every: float | None = None,
        show_label: bool | None = None,
        container: bool = True,
        scale: int | None = None,
        min_width: int = 160,
        visible: bool = True,
        elem_id: str | None = None,
        elem_classes: list[str] | str | None = None,
        render: bool = True,
        key: int | str | None = None,
    ):
        """
        Parameters:
            value: Tuple of base image and list of (annotation, label) pairs.
            format: Format used to save images before it is returned to the front end, such as 'jpeg' or 'png'. This parameter only takes effect when the base image is returned from the prediction function as a numpy array or a PIL Image. The format should be supported by the PIL library.
            show_legend: If True, will show a legend of the annotations.
            height: The height of the image, specified in pixels if a number is passed, or in CSS units if a string is passed.
            width: The width of the image, specified in pixels if a number is passed, or in CSS units if a string is passed.
            color_map: A dictionary mapping labels to colors. The colors must be specified as hex codes.
            label: 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.
            every: If `value` is a callable, run the function 'every' number of seconds while the client connection is open. Has no effect otherwise. The event can be accessed (e.g. to cancel it) via this component's .load_event attribute.
            show_label: if True, will display label.
            container: If True, will place the component in a container - providing some extra padding around the border.
            scale: Relative width compared to adjacent Components in a Row. For example, if Component A has scale=2, and Component B has scale=1, A will be twice as wide as B. Should be an integer.
            min_width: 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.
            visible: If False, component will be hidden.
            elem_id: 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: 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: 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.
            key: if assigned, will be used to assume identity across a re-render. Components that have the same key across a re-render will have their value preserved.
        """
        self.format = format
        self.show_legend = show_legend
        self.height = height
        self.width = width
        self.color_map = color_map
        super().__init__(
            label=label,
            every=every,
            show_label=show_label,
            container=container,
            scale=scale,
            min_width=min_width,
            visible=visible,
            elem_id=elem_id,
            elem_classes=elem_classes,
            render=render,
            key=key,
            value=value,
        )

    def preprocess(
        self, payload: AnnotatedImageData | None
    ) -> tuple[str, list[tuple[str, str]]] | None:
        """
        Parameters:
            payload: Dict of base image and list of annotations.
        Returns:
            Passes its value as a `tuple` consisting of a `str` filepath to a base image and `list` of annotations. Each annotation itself is `tuple` of a mask (as a `str` filepath to image) and a `str` label.
        """
        if payload is None:
            return None
        base_img = payload.image.path
        annotations = [(a.image.path, a.label) for a in payload.annotations]
        return (base_img, annotations)

    def postprocess(
        self,
        value: (
            tuple[
                np.ndarray | PIL.Image.Image | str,
                list[tuple[np.ndarray | tuple[int, int, int, int], str]],
            ]
            | None
        ),
    ) -> AnnotatedImageData | None:
        """
        Parameters:
            value: Expects a a tuple of a base image and list of annotations: a `tuple[Image, list[Annotation]]`. The `Image` itself can be `str` filepath, `numpy.ndarray`, or `PIL.Image`. Each `Annotation` is a `tuple[Mask, str]`. The `Mask` can be either a `tuple` of 4 `int`'s representing the bounding box coordinates (x1, y1, x2, y2), or 0-1 confidence mask in the form of a `numpy.ndarray` of the same shape as the image, while the second element of the `Annotation` tuple is a `str` label.
        Returns:
            Tuple of base image file and list of annotations, with each annotation a two-part tuple where the first element image path of the mask, and the second element is the label.
        """
        if value is None:
            return None
        base_img = value[0]
        if isinstance(base_img, str):
            if client_utils.is_http_url_like(base_img):
                base_img = processing_utils.save_url_to_cache(
                    base_img, cache_dir=self.GRADIO_CACHE
                )
            base_img_path = base_img
            base_img = np.array(PIL.Image.open(base_img))
        elif isinstance(base_img, np.ndarray):
            base_file = processing_utils.save_img_array_to_cache(
                base_img, cache_dir=self.GRADIO_CACHE, format=self.format
            )
            base_img_path = str(utils.abspath(base_file))
        elif isinstance(base_img, PIL.Image.Image):
            base_file = processing_utils.save_pil_to_cache(
                base_img, cache_dir=self.GRADIO_CACHE, format=self.format
            )
            base_img_path = str(utils.abspath(base_file))
            base_img = np.array(base_img)
        else:
            raise ValueError(
                "AnnotatedImage only accepts filepaths, PIL images or numpy arrays for the base image."
            )

        sections = []
        color_map = self.color_map or {}

        def hex_to_rgb(value):
            value = value.lstrip("#")
            lv = len(value)
            return [int(value[i : i + lv // 3], 16) for i in range(0, lv, lv // 3)]

        for mask, label in value[1]:
            mask_array = np.zeros((base_img.shape[0], base_img.shape[1]))
            if isinstance(mask, np.ndarray):
                mask_array = mask
            else:
                x1, y1, x2, y2 = mask
                border_width = 3
                mask_array[y1:y2, x1:x2] = 0.5
                mask_array[y1:y2, x1 : x1 + border_width] = 1
                mask_array[y1:y2, x2 - border_width : x2] = 1
                mask_array[y1 : y1 + border_width, x1:x2] = 1
                mask_array[y2 - border_width : y2, x1:x2] = 1

            if label in color_map:
                rgb_color = hex_to_rgb(color_map[label])
            else:
                rgb_color = [255, 0, 0]
            colored_mask = np.zeros((base_img.shape[0], base_img.shape[1], 4))
            solid_mask = np.copy(mask_array)
            solid_mask[solid_mask > 0] = 1

            colored_mask[:, :, 0] = rgb_color[0] * solid_mask
            colored_mask[:, :, 1] = rgb_color[1] * solid_mask
            colored_mask[:, :, 2] = rgb_color[2] * solid_mask
            colored_mask[:, :, 3] = mask_array * 255

            colored_mask_img = PIL.Image.fromarray((colored_mask).astype(np.uint8))

            # RGBA does not support transparency
            mask_file = processing_utils.save_pil_to_cache(
                colored_mask_img, cache_dir=self.GRADIO_CACHE, format="png"
            )
            mask_file_path = str(utils.abspath(mask_file))
            sections.append(
                Annotation(image=FileData(path=mask_file_path), label=label)
            )

        return AnnotatedImageData(
            image=FileData(path=base_img_path),
            annotations=sections,
        )

    def example_payload(self) -> Any:
        return {
            "image": file(
                "https://raw.githubusercontent.com/gradio-app/gradio/main/test/test_files/bus.png"
            ),
            "annotations": [],
        }

    def example_value(self) -> Any:
        return (
            "https://raw.githubusercontent.com/gradio-app/gradio/main/test/test_files/bus.png",
            [([0, 0, 100, 100], "bus")],
        )

    
    def select(self,
        fn: Callable | None,
        inputs: Component | Sequence[Component] | set[Component] | None = None,
        outputs: Component | Sequence[Component] | None = None,
        api_name: str | None | Literal[False] = None,
        scroll_to_output: bool = False,
        show_progress: Literal["full", "minimal", "hidden"] = "full",
        queue: bool | None = None,
        batch: bool = False,
        max_batch_size: int = 4,
        preprocess: bool = True,
        postprocess: bool = True,
        cancels: dict[str, Any] | list[dict[str, Any]] | None = None,
        every: float | None = None,
        trigger_mode: Literal["once", "multiple", "always_last"] | None = None,
        js: str | None = None,
        concurrency_limit: int | None | Literal["default"] = "default",
        concurrency_id: str | None = None,
        show_api: bool = True) -> Dependency:
        """
        Parameters:
            fn: the function to call when this event is triggered. Often a machine learning model's prediction function. Each parameter of the function corresponds to one input component, and the function should return a single value or a tuple of values, with each element in the tuple corresponding to one output component.
            inputs: List of gradio.components to use as inputs. If the function takes no inputs, this should be an empty list.
            outputs: List of gradio.components to use as outputs. If the function returns no outputs, this should be an empty list.
            api_name: Defines how the endpoint appears in the API docs. Can be a string, None, or False. If False, the endpoint will not be exposed in the api docs. If set to None, the endpoint will be exposed in the api docs as an unnamed endpoint, although this behavior will be changed in Gradio 4.0. If set to a string, the endpoint will be exposed in the api docs with the given name.
            scroll_to_output: If True, will scroll to output component on completion
            show_progress: If True, will show progress animation while pending
            queue: If True, will place the request on the queue, if the queue has been enabled. If False, will not put this event on the queue, even if the queue has been enabled. If None, will use the queue setting of the gradio app.
            batch: If True, then the function should process a batch of inputs, meaning that it should accept a list of input values for each parameter. The lists should be of equal length (and be up to length `max_batch_size`). The function is then *required* to return a tuple of lists (even if there is only 1 output component), with each list in the tuple corresponding to one output component.
            max_batch_size: Maximum number of inputs to batch together if this is called from the queue (only relevant if batch=True)
            preprocess: If False, will not run preprocessing of component data before running 'fn' (e.g. leaving it as a base64 string if this method is called with the `Image` component).
            postprocess: If False, will not run postprocessing of component data before returning 'fn' output to the browser.
            cancels: A list of other events to cancel when this listener is triggered. For example, setting cancels=[click_event] will cancel the click_event, where click_event is the return value of another components .click method. Functions that have not yet run (or generators that are iterating) will be cancelled, but functions that are currently running will be allowed to finish.
            every: Run this event 'every' number of seconds while the client connection is open. Interpreted in seconds.
            trigger_mode: If "once" (default for all events except `.change()`) would not allow any submissions while an event is pending. If set to "multiple", unlimited submissions are allowed while pending, and "always_last" (default for `.change()` and `.key_up()` events) would allow a second submission after the pending event is complete.
            js: Optional frontend js method to run before running 'fn'. Input arguments for js method are values of 'inputs' and 'outputs', return should be a list of values for output components.
            concurrency_limit: If set, this is the maximum number of this event that can be running simultaneously. Can be set to None to mean no concurrency_limit (any number of this event can be running simultaneously). Set to "default" to use the default concurrency limit (defined by the `default_concurrency_limit` parameter in `Blocks.queue()`, which itself is 1 by default).
            concurrency_id: If set, this is the id of the concurrency group. Events with the same concurrency_id will be limited by the lowest set concurrency_limit.
            show_api: whether to show this event in the "view API" page of the Gradio app, or in the ".view_api()" method of the Gradio clients. Unlike setting api_name to False, setting show_api to False will still allow downstream apps as well as the Clients to use this event. If fn is None, show_api will automatically be set to False.
        """
        ...