File size: 18,159 Bytes
97ed94f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b864380
63d71fd
 
b864380
 
97ed94f
87a479d
63d71fd
 
 
6eb6415
 
 
 
63d71fd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92b9418
63d71fd
 
 
 
 
 
 
 
92b9418
b864380
 
 
 
 
97ed94f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
85562af
b864380
 
 
 
97ed94f
 
 
 
 
 
 
b864380
 
 
 
97ed94f
 
 
 
 
b864380
 
87a479d
b864380
97ed94f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
85562af
b864380
87a479d
97ed94f
 
 
 
 
 
 
 
 
b864380
 
 
 
 
 
97ed94f
 
 
 
 
 
 
 
b864380
933222d
 
b864380
 
 
87a479d
 
 
 
 
 
 
b864380
 
 
97ed94f
 
 
 
 
 
 
 
 
 
b864380
 
 
 
 
 
97ed94f
 
 
 
 
 
 
b864380
 
 
97ed94f
 
 
 
 
 
 
 
b864380
 
 
 
 
 
 
87a479d
b864380
 
87a479d
b864380
 
 
87a479d
933222d
85562af
 
97ed94f
 
 
 
 
 
87a479d
97ed94f
 
87a479d
97ed94f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
85562af
b864380
97ed94f
b864380
92b9418
8b391b7
1366c9e
87a479d
1366c9e
87a479d
b864380
 
1366c9e
8b391b7
b864380
933222d
 
97ed94f
 
 
 
 
 
 
 
 
87a479d
b864380
87a479d
 
b864380
 
87a479d
 
 
b864380
97ed94f
 
 
 
 
 
 
 
 
b864380
 
 
 
 
 
 
 
 
1366c9e
b864380
 
 
 
 
97ed94f
 
c4cb2d4
97ed94f
 
 
 
 
 
b864380
e42b13d
b864380
e42b13d
 
 
 
 
 
 
c4cb2d4
 
b864380
97ed94f
87a479d
97ed94f
 
 
 
 
 
 
 
 
87a479d
 
 
 
 
 
 
97ed94f
 
 
 
 
 
 
 
 
 
 
87a479d
b864380
 
97ed94f
 
 
 
 
 
 
 
 
 
85562af
933222d
b864380
 
8b391b7
 
 
1366c9e
b864380
 
 
 
8b391b7
 
b864380
c4cb2d4
87a479d
 
 
933222d
87a479d
b864380
 
 
1366c9e
87a479d
 
933222d
b864380
 
1366c9e
87a479d
 
 
e42b13d
b864380
 
999040d
87a479d
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
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
"""
This module integrates real-time object detection into live YouTube streams using the YOLO (You Only Look Once) model, and provides an interactive user interface through Gradio. It is designed to allow users to search for live YouTube streams and apply object detection to these streams in real time.

Main Features:
- Search for live YouTube streams using specific queries.
- Retrieve live stream URLs using the Streamlink library.
- Perform real-time object detection on live streams using the YOLO model.
- Display the live stream and object detection results through a Gradio interface.

The module comprises several key components:
- `SearchFilter`: An enumeration for YouTube search filters.
- `SearchService`: A service class to search for YouTube videos and retrieve live stream URLs.
- `LiveYouTubeObjectDetector`: The main class integrating the YOLO model and Gradio UI, handling the entire workflow of searching, streaming, and object detection.

Dependencies:
- cv2 (OpenCV): Used for image processing tasks.
- Gradio: Provides the interactive web-based user interface.
- innertube, streamlink: Used for interacting with YouTube and retrieving live stream data.
- numpy: Utilized for numerical operations on image data.
- PIL (Pillow): A Python Imaging Library for opening, manipulating, and saving images.
- ultralytics YOLO: The YOLO model implementation for object detection.

Usage:
Run this file to launch the Gradio interface, which allows users to input search queries for YouTube live streams, select a stream, and perform object detection on the selected live stream.

"""
import logging
import os
import subprocess
import sys
from enum import Enum
from typing import Any, Dict, List, Optional, Tuple

import requests


# HOTFIX: https://github.com/aws/aws-cli/issues/6130#issuecomment-829659803
os.environ['AWS_EC2_METADATA_DISABLED'] = 'true'


def install_requirements():
    requirements_url = "https://raw.githubusercontent.com/aai521-group6/project/main/requirements.txt"
    response = requests.get(requirements_url)
    if response.status_code == 200:
        with open("requirements.txt", "wb") as file:
            file.write(response.content)
        subprocess.check_call([sys.executable, "-m", "pip", "install", "-r", "requirements.txt"])
    else:
        raise Exception("Failed to download requirements.txt")


try:
    import cv2
    import gradio as gr
    import innertube
    import numpy as np
    import streamlink
    from PIL import Image
    from ultralytics import YOLO
except ImportError:
    install_requirements()
    import cv2
    import gradio as gr
    import innertube
    import numpy as np
    import streamlink
    from PIL import Image
    from ultralytics import YOLO

logging.basicConfig(stream=sys.stderr, level=logging.DEBUG)


class SearchFilter(Enum):
    """
    An enumeration for specifying different types of YouTube search filters.

    This Enum class is used to define filters for categorizing YouTube search
    results into either live or regular video content. It is utilized in
    conjunction with the `SearchService` class to refine YouTube searches
    based on the type of content being sought.

    Attributes:
        LIVE (str): Represents the filter code for live video content on YouTube.
        VIDEO (str): Represents the filter code for regular, non-live video content on YouTube.

    Each attribute consists of a tuple where the first element is the filter code
    used in YouTube search queries, and the second element is a human-readable
    string describing the filter.
    """

    LIVE = ("EgJAAQ%3D%3D", "Live")
    VIDEO = ("EgIQAQ%3D%3D", "Video")

    def __init__(self, code, human_readable):
        """Initializes the SearchFilter with a code and a human-readable string.

        :param code: The filter code used in YouTube search queries.
        :type code: str
        :param human_readable: A human-readable representation of the filter.
        :type human_readable: str
        """
        self.code = code
        self.human_readable = human_readable

    def __str__(self):
        """Returns the human-readable representation of the filter.

        :return: The human-readable representation of the filter.
        :rtype: str
        """
        return self.human_readable


class SearchService:
    """
    SearchService provides functionality to search for YouTube videos using the
    InnerTube API and retrieve live stream URLs using the Streamlink library.

    This service allows filtering search results to either live or regular video
    content and parsing the search response to extract relevant video information.
    It also constructs YouTube URLs for given video IDs and retrieves the best
    available stream URL for live YouTube videos.

    Methods:
        search: Searches YouTube for videos matching a query and filter.
        parse: Parses raw search response data into a list of video details.
        _search: Performs a YouTube search with the given query and filter.
        get_youtube_url: Constructs a YouTube URL for a given video ID.
        get_stream: Retrieves the stream URL for a given YouTube video URL.
    """

    @staticmethod
    def search(query: Optional[str], filter: SearchFilter = SearchFilter.VIDEO):
        """Searches YouTube for videos matching the given query and filter.

        :param query: The search query.
        :type query: Optional[str]
        :param filter: The search filter to apply.
        :type filter: SearchFilter
        :return: A list of search results, each a dictionary with video details.
        :rtype: List[Dict[str, Any]]
        """
        client = innertube.InnerTube("WEB", "2.20230920.00.00")
        response = SearchService._search(query, filter)
        results = SearchService.parse(response)
        return results

    @staticmethod
    def parse(data: Dict[str, Any]) -> List[Dict[str, str]]:
        """Parses the raw search response data into a list of video details.

        :param data: The raw search response data from YouTube.
        :type data: Dict[str, Any]
        :return: A list of parsed video details.
        :rtype: List[Dict[str, str]]
        """
        results = []
        contents = data["contents"]["twoColumnSearchResultsRenderer"]["primaryContents"]["sectionListRenderer"]["contents"]
        items = contents[0]["itemSectionRenderer"]["contents"] if contents else []
        for item in items:
            if "videoRenderer" in item:
                renderer = item["videoRenderer"]
                results.append(
                    {
                        "video_id": renderer["videoId"],
                        "thumbnail_url": renderer["thumbnail"]["thumbnails"][-1]["url"],
                        "title": "".join(run["text"] for run in renderer["title"]["runs"]),
                    }
                )
        return results

    @staticmethod
    def _search(query: Optional[str] = None, filter: SearchFilter = SearchFilter.VIDEO) -> Dict[str, Any]:
        """Performs a YouTube search with the given query and filter.

        :param query: The search query.
        :type query: Optional[str]
        :param filter: The search filter to apply.
        :type filter: SearchFilter
        :return: The raw search response data from YouTube.
        :rtype: Dict[str, Any]
        """
        client = innertube.InnerTube("WEB", "2.20230920.00.00")
        response = client.search(query=query, params=filter.code if filter else None)
        return response

    @staticmethod
    def get_youtube_url(video_id: str) -> str:
        """Constructs a YouTube URL for the given video ID.

        :param video_id: The ID of the YouTube video.
        :type video_id: str
        :return: The YouTube URL for the video.
        :rtype: str
        """
        return f"https://www.youtube.com/watch?v={video_id}"

    @staticmethod
    def get_stream(youtube_url: str) -> Optional[str]:
        """Retrieves the stream URL for a given YouTube video URL.

        :param youtube_url: The URL of the YouTube video.
        :type youtube_url: str
        :return: The stream URL if available, otherwise None.
        :rtype: Optional[str]
        """
        try:
            session = streamlink.Streamlink()
            streams = session.streams(youtube_url)
            if streams:
                best_stream = streams.get("best")
                return best_stream.url if best_stream else None
            else:
                gr.Warning(f"No streams found for: {youtube_url}")
                return None
        except Exception as e:
            gr.Error(f"An error occurred while getting stream: {e}")
            logging.warning(f"An error occurred: {e}")
            return None


INITIAL_STREAMS = SearchService.search("world live cams", SearchFilter.LIVE)


class LiveYouTubeObjectDetector:
    """
    LiveYouTubeObjectDetector is a class that integrates object detection into live YouTube streams.
    It uses the YOLO (You Only Look Once) model to detect objects in video frames captured from live streams.
    The class also provides a Gradio interface for users to interact with the object detection system,
    allowing them to search for live streams, view them, and detect objects in real-time.

    The class handles the retrieval of live stream URLs, frame capture from the streams, object detection
    on the frames, and updating the Gradio interface with the results.

    Attributes:
        model (YOLO): The YOLO model used for object detection.
        streams (list): A list of dictionaries containing information about the current live streams.
        gallery (gr.Gallery): A Gradio gallery widget to display live stream thumbnails.
        search_input (gr.Textbox): A Gradio textbox for inputting search queries.
        stream_input (gr.Textbox): A Gradio textbox for inputting a specific live stream URL.
        annotated_image (gr.AnnotatedImage): A Gradio annotated image widget to display detection results.
        search_button (gr.Button): A Gradio button to initiate a new search for live streams.
        submit_button (gr.Button): A Gradio button to start object detection on a specified live stream.
        page_title (gr.HTML): A Gradio HTML widget to display the page title.

    Methods:
        detect_objects: Detects objects in a live YouTube stream given its URL.
        get_frame: Captures a frame from a live stream URL.
        annotate: Annotates a frame with detected objects.
        create_black_image: Creates a black placeholder image.
        get_live_streams: Searches for live streams based on a query.
        render: Sets up and launches the Gradio interface.
    """

    def __init__(self):
        """Initializes the LiveYouTubeObjectDetector with YOLO model and UI components."""
        logging.getLogger().setLevel(logging.DEBUG)
        self.model = YOLO("yolov8x.pt")
        self.streams = INITIAL_STREAMS

        # Gradio UI
        initial_gallery_items = [(stream["thumbnail_url"], stream["title"]) for stream in self.streams]
        self.gallery = gr.Gallery(label="Live YouTube Videos", value=initial_gallery_items, show_label=True, columns=[4], rows=[5], object_fit="contain", height="auto", allow_preview=False)
        self.search_input = gr.Textbox(label="Search Live YouTube Videos")
        self.stream_input = gr.Textbox(label="URL of Live YouTube Video")
        self.annotated_image = gr.AnnotatedImage(show_label=False)
        self.search_button = gr.Button("Search", size="lg")
        self.submit_button = gr.Button("Detect Objects", variant="primary", size="lg")
        self.page_title = gr.HTML("<center><h1><b>Object Detection in Live YouTube Streams</b></h1></center>")

    def detect_objects(self, url: str) -> Tuple[Image.Image, List[Tuple[Tuple[int, int, int, int], str]]]:
        """
        Detects objects in the given live YouTube stream URL.

        :param url: The URL of the live YouTube video.
        :type url: str
        :return: A tuple containing the annotated image and a list of annotations.
        :rtype: Tuple[Image.Image, List[Tuple[Tuple[int, int, int, int], str]]]
        """
        stream_url = SearchService.get_stream(url)
        if not stream_url:
            gr.Error(f"Unable to find a stream for: {stream_url}")
            return self.create_black_image(), []
        frame = self.get_frame(stream_url)
        if frame is None:
            gr.Error(f"Unable to capture frame for: {stream_url}")
            return self.create_black_image(), []
        return self.annotate(frame)

    def get_frame(self, stream_url: str) -> Optional[np.ndarray]:
        """
        Captures a frame from the given live stream URL.

        :param stream_url: The URL of the live stream.
        :type stream_url: str
        :return: The captured frame as a numpy array, or None if capture fails.
        :rtype: Optional[np.ndarray]
        """
        if not stream_url:
            return None
        try:
            cap = cv2.VideoCapture(stream_url)
            ret, frame = cap.read()
            cap.release()
            if ret:
                return cv2.resize(frame, (1920, 1080))
            else:
                logging.warning("Unable to process the HLS stream with cv2.VideoCapture.")
                return None
        except Exception as e:
            logging.warning(f"An error occurred while capturing the frame: {e}")
            return None

    def annotate(self, frame: np.ndarray) -> Tuple[Image.Image, List[Tuple[Tuple[int, int, int, int], str]]]:
        """
        Annotates the given frame with detected objects and their bounding boxes.

        :param frame: The frame to be annotated.
        :type frame: np.ndarray
        :return: A tuple of the annotated PIL image and list of annotations.
        :rtype: Tuple[Image.Image, List[Tuple[Tuple[int, int, int, int], str]]]
        """
        frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
        predictions = self.model.predict(frame_rgb)
        annotations = []
        result = predictions._images_prediction_lst[0]

        for bbox, label in zip(result.prediction.bboxes_xyxy, result.prediction.labels):
            x1, y1, x2, y2 = bbox
            class_name = result.class_names[int(label)]
            bbox_coords = (int(x1), int(y1), int(x2), int(y2))
            annotations.append((bbox_coords, class_name))

        return Image.fromarray(frame_rgb), annotations

    @staticmethod
    def create_black_image():
        """
        Creates a black image of fixed dimensions.

        This method generates a black image that can be used as a placeholder or background.
        It is particularly useful in cases where no valid frame is available for processing.

        :return: A black image as a numpy array.
        :rtype: np.ndarray
        """
        black_image = np.zeros((1080, 1920, 3), dtype=np.uint8)
        pil_black_image = Image.fromarray(black_image)
        cv2_black_image = cv2.cvtColor(np.array(pil_black_image), cv2.COLOR_RGB2BGR)
        return cv2_black_image

    @staticmethod
    def get_live_streams(query=""):
        """
        Searches for live streams on YouTube based on the given query.

        This method utilizes the SearchService to find live YouTube streams. If no query is
        provided, it defaults to searching for 'world live cams'.

        :param query: The search query for live streams, defaults to an empty string.
        :type query: str, optional
        :return: A list of dictionaries containing information about each live stream.
        :rtype: List[Dict[str, str]]
        """
        return SearchService.search(query if query else "world live cams", SearchFilter.LIVE)

    def render(self):
        """
        Sets up and launches the Gradio interface for the application.

        This method creates the Gradio UI elements and defines the behavior of the application.
        It includes the setup of interactive widgets like galleries, textboxes, and buttons,
        and defines the actions triggered by user interactions with these widgets.

        The Gradio interface allows users to search for live YouTube streams, select a stream,
        and run object detection on the selected live stream.
        """
        with gr.Blocks(title="Object Detection in Live YouTube Streams", css="footer {visibility: hidden}", analytics_enabled=False) as app:
            self.page_title.render()
            with gr.Column():
                with gr.Group():
                    with gr.Row():
                        self.stream_input.render()
                        self.submit_button.render()
                self.annotated_image.render()
            with gr.Group():
                with gr.Row():
                    self.search_input.render()
                    self.search_button.render()
            with gr.Row():
                self.gallery.render()

            @self.gallery.select(inputs=None, outputs=[self.annotated_image, self.stream_input], scroll_to_output=True)
            def detect_objects_from_gallery_item(evt: gr.SelectData):
                if evt.index is not None and evt.index < len(self.streams):
                    selected_stream = self.streams[evt.index]
                    stream_url = SearchService.get_youtube_url(selected_stream["video_id"])
                    frame_output = self.detect_objects(stream_url)
                    return frame_output, stream_url
                return None, ""

            @self.search_button.click(inputs=[self.search_input], outputs=[self.gallery])
            def search_live_streams(query):
                self.streams = self.get_live_streams(query)
                gallery_items = [(stream["thumbnail_url"], stream["title"]) for stream in self.streams]
                return gallery_items

            @self.submit_button.click(inputs=[self.stream_input], outputs=[self.annotated_image])
            def detect_objects_from_url(url):
                return self.detect_objects(url)

        return app.queue().launch(show_api=False, debug=True)


if __name__ == "__main__":
    LiveYouTubeObjectDetector().render()