File size: 23,199 Bytes
a27b4e4 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 |
import os
import cv2
import pickle
import numpy as np
from keras.models import load_model
from moviepy.editor import VideoFileClip
import librosa
from sklearn.preprocessing import LabelEncoder
import xml.etree.ElementTree as ET
import random
class VideoAudioFeatureExtractor:
def __init__(self, video_path, output_path):
self.video_path = video_path
self.output_path = output_path
self.num_frames = 15
self.height = 224
self.width = 224
self.channels = 6
self.audio_feature_dim = 20
self.num_actions = 3
self.classification_model = load_model("EdAi-gamestyle.h5")
with open("label_encoder.pkl", "rb") as f:
self.label_encoder = pickle.load(f)
with open("clip_items.pkl", "rb") as f:
self.all_clip_items = pickle.load(f)
def generate_clipitem_id(self, video_name, num_digits=10):
random_number = ''.join(random.choices('0123456789', k=num_digits))
return f"{video_name}{random_number}"
def combine_features(self, video_features, audio_features):
audio_features_resized = cv2.resize(audio_features, (self.height, self.width))
audio_features_reshaped = np.repeat(audio_features_resized[:, :, np.newaxis], 3, axis=2)
audio_features_reshaped = np.repeat(audio_features_reshaped[np.newaxis, :, :, :], self.num_frames, axis=0)
combined_features = np.concatenate([video_features, audio_features_reshaped], axis=-1)
return combined_features
def extract_audio(self, video_path):
clip = VideoFileClip(video_path)
audio = clip.audio
audio_signal = audio.to_soundarray(fps=48000)
audio_signal = audio_signal.astype(np.float32)
if len(audio_signal.shape) == 2:
audio_signal = librosa.to_mono(audio_signal.T)
return audio_signal
def process_audio(self, audio_signal_segment):
mfccs = librosa.feature.mfcc(y=audio_signal_segment, sr=48000, n_mfcc=self.audio_feature_dim)
mfccs_fixed_length = librosa.util.fix_length(mfccs, size=self.num_frames)
return mfccs_fixed_length.T
def preprocess_frame(self, frame):
frame_resized = cv2.resize(frame, (self.width, self.height))
frame_normalized = (frame_resized / 255.0).astype(np.float32)
return frame_normalized
def extract_clip(self, video, audio_features, decision):
start_frame = int(decision["in"])
end_frame = int(decision["out"])
video.set(cv2.CAP_PROP_POS_FRAMES, start_frame)
clip_frames = []
for i in range(start_frame, end_frame, 2):
ret, frame = video.read()
if not ret:
break
frame_processed = self.preprocess_frame(frame)
clip_frames.append(frame_processed)
while len(clip_frames) < self.num_frames:
clip_frames.append(np.zeros((self.height, self.width, 3), dtype=np.float32))
clip_frames = np.array(clip_frames[:self.num_frames])
return clip_frames, audio_features
def create_audio_track(self, clip_item_ids, pan_value, link_ids, video_clip_items, video_file_elements):
track = ET.Element("track")
for idx, clipitem_id in enumerate(clip_item_ids):
clipitem_element = ET.Element("clipitem", id=clipitem_id)
video_clip_item = video_clip_items[idx]
in_value = video_clip_item['in']
out_value = video_clip_item['out']
start_value = video_clip_item['start']
end_value = video_clip_item['end']
ET.SubElement(clipitem_element, "name").text = video_clip_item['name']
ET.SubElement(clipitem_element, "duration").text = video_clip_item['duration']
rate_elem = ET.SubElement(clipitem_element, "rate")
ET.SubElement(rate_elem, "ntsc").text = "TRUE"
ET.SubElement(rate_elem, "timebase").text = "30"
ET.SubElement(clipitem_element, "in").text = in_value
ET.SubElement(clipitem_element, "out").text = out_value
ET.SubElement(clipitem_element, "start").text = start_value
ET.SubElement(clipitem_element, "end").text = end_value
ET.SubElement(clipitem_element, "masterclipid").text = os.path.basename(self.video_path)
sourcetrack = ET.SubElement(clipitem_element, "sourcetrack")
ET.SubElement(sourcetrack, "mediatype").text = "audio"
ET.SubElement(sourcetrack, "trackindex").text = str(idx + 1)
video_file_elem = video_clip_item.get("file")
if video_file_elem is not None:
ET.SubElement(clipitem_element, "file", id=video_file_elem.get("id"))
else:
ET.SubElement(clipitem_element, "file", id=os.path.basename(self.video_path).split('.')[0])
filter_elem = ET.SubElement(clipitem_element, "filter")
effect_elem = ET.SubElement(filter_elem, "effect")
ET.SubElement(effect_elem, "name").text = "Audio Levels"
ET.SubElement(effect_elem, "effectid").text = "audiolevels"
ET.SubElement(effect_elem, "effectcategory").text = "audiolevels"
ET.SubElement(effect_elem, "effecttype").text = "audiolevels"
ET.SubElement(effect_elem, "mediatype").text = "audio"
parameter_elem = ET.SubElement(effect_elem, "parameter")
ET.SubElement(parameter_elem, "name").text = "Level"
ET.SubElement(parameter_elem, "parameterid").text = "level"
ET.SubElement(parameter_elem, "valuemin").text = "0"
ET.SubElement(parameter_elem, "valuemax").text = "3.98109"
ET.SubElement(parameter_elem, "value").text = "1"
filter_elem = ET.SubElement(clipitem_element, "filter")
effect_elem = ET.SubElement(filter_elem, "effect")
ET.SubElement(effect_elem, "name").text = "Audio Pan"
ET.SubElement(effect_elem, "effectid").text = "audiopan"
ET.SubElement(effect_elem, "effectcategory").text = "audiopan"
ET.SubElement(effect_elem, "effecttype").text = "audiopan"
ET.SubElement(effect_elem, "mediatype").text = "audio"
parameter_elem = ET.SubElement(effect_elem, "parameter")
ET.SubElement(parameter_elem, "name").text = "Pan"
ET.SubElement(parameter_elem, "parameterid").text = "pan"
ET.SubElement(parameter_elem, "valuemin").text = "-1"
ET.SubElement(parameter_elem, "valuemax").text = "1"
ET.SubElement(parameter_elem, "value").text = str(pan_value)
for link_id in link_ids[idx]:
link_elem = ET.SubElement(clipitem_element, "link")
ET.SubElement(link_elem, "linkclipref").text = link_id
track.append(clipitem_element)
self.adjust_clipitem_start_end(track.findall('clipitem'))
return track
def create_clip_structure(self, video_name, video_path, total_duration):
clip = ET.Element("clip", id=video_name)
ET.SubElement(clip, "updatebehavior").text = "add"
ET.SubElement(clip, "name").text = video_name
ET.SubElement(clip, "duration").text = str(total_duration)
rate = ET.SubElement(clip, "rate")
ET.SubElement(rate, "ntsc").text = "TRUE"
ET.SubElement(rate, "timebase").text = "30"
ET.SubElement(clip, "in").text = "-1"
ET.SubElement(clip, "out").text = "-1"
ET.SubElement(clip, "masterclipid").text = video_name
ET.SubElement(clip, "ismasterclip").text = "TRUE"
# Logging info
logginginfo = ET.SubElement(clip, "logginginfo")
ET.SubElement(logginginfo, "scene")
ET.SubElement(logginginfo, "shottake")
ET.SubElement(logginginfo, "lognote")
ET.SubElement(logginginfo, "good").text = "FALSE"
# Labels
labels = ET.SubElement(clip, "labels")
ET.SubElement(labels, "label2")
# Comments
comments = ET.SubElement(clip, "comments")
ET.SubElement(comments, "mastercomment1")
ET.SubElement(comments, "mastercomment2")
ET.SubElement(comments, "mastercomment3")
ET.SubElement(comments, "mastercomment4")
# Media
media = ET.SubElement(clip, "media")
video = ET.SubElement(media, "video")
track = ET.SubElement(video, "track")
clipitem = ET.SubElement(track, "clipitem", id=f"{video_name}1")
ET.SubElement(clipitem, "name").text = video_name
ET.SubElement(clipitem, "duration").text = str(total_duration)
rate = ET.SubElement(clipitem, "rate")
ET.SubElement(rate, "ntsc").text = "TRUE"
ET.SubElement(rate, "timebase").text = "30"
ET.SubElement(clipitem, "in").text = "0"
ET.SubElement(clipitem, "out").text = str(total_duration)
ET.SubElement(clipitem, "start").text = "0"
ET.SubElement(clipitem, "end").text = str(total_duration)
ET.SubElement(clipitem, "pixelaspectratio").text = "Square"
ET.SubElement(clipitem, "anamorphic").text = "FALSE"
ET.SubElement(clipitem, "alphatype").text = "none"
ET.SubElement(clipitem, "masterclipid").text = video_name
# File
file_id = video_name.split('.')[0] # Extract filename without extension
file = ET.SubElement(clipitem, "file")
file.text = f'file id="{file_id}"' # Set the text of the <file> tag to the file ID without the extension
ET.SubElement(file, "name").text = video_name
ET.SubElement(file, "pathurl").text = f"file://localhost/{video_path}"
rate = ET.SubElement(file, "rate")
ET.SubElement(rate, "timebase").text = "30"
ET.SubElement(rate, "ntsc").text = "TRUE"
ET.SubElement(file, "duration").text = str(total_duration)
timecode = ET.SubElement(file, "timecode")
ET.SubElement(timecode, "string").text = "00:00:00:00"
ET.SubElement(timecode, "frame").text = "0"
ET.SubElement(timecode, "rate").text = "30"
ET.SubElement(timecode, "displayformat").text = "NDF"
media = ET.SubElement(file, "media")
video = ET.SubElement(media, "video")
audio = ET.SubElement(media, "audio")
# Audio
for i in range(2): # Assuming there are 2 audio tracks as per the example
track = ET.SubElement(audio, "track")
clipitem = ET.SubElement(track, "clipitem", id=f"{video_name}{i + 2}")
ET.SubElement(clipitem, "name").text = video_name
ET.SubElement(clipitem, "duration").text = str(total_duration)
rate = ET.SubElement(clipitem, "rate")
ET.SubElement(rate, "ntsc").text = "TRUE"
ET.SubElement(rate, "timebase").text = "30"
ET.SubElement(clipitem, "in").text = "0"
ET.SubElement(clipitem, "out").text = str(total_duration)
ET.SubElement(clipitem, "start").text = "0"
ET.SubElement(clipitem, "end").text = str(total_duration)
ET.SubElement(clipitem, "masterclipid").text = video_name
file_id = video_name.split('.')[0] # Extract filename without extension
file_elem = ET.SubElement(clipitem, "file", id=file_id)
ET.SubElement(file_elem, "name").text = video_name
ET.SubElement(file_elem, "pathurl").text = f"file://localhost/{video_path}"
sourcetrack = ET.SubElement(clipitem, "sourcetrack")
ET.SubElement(sourcetrack, "mediatype").text = "audio"
ET.SubElement(sourcetrack, "trackindex").text = str(i + 1)
return clip
def adjust_clipitem_start_end(self, clipitems):
last_end = 0
for clipitem in clipitems:
in_val = int(clipitem.find('in').text)
out_val = int(clipitem.find('out').text)
clipitem.find('start').text = str(last_end)
clipitem.find('end').text = str(last_end + (out_val - in_val))
last_end = int(clipitem.find('end').text)
def dict_to_xml(self, tag, dictionary):
attributes = dictionary.pop('attributes', {})
if tag == "clipitem" or tag == "file":
attributes["id"] = dictionary.pop("id", None)
for key, val in attributes.items():
if val is None:
attributes[key] = "none"
elem = ET.Element(tag, **attributes)
for key, val in dictionary.items():
if val is None:
val = "none"
if isinstance(val, dict):
child = self.dict_to_xml(key, val)
elem.append(child)
elif isinstance(val, list):
for item in val:
child = self.dict_to_xml(key, item)
elem.append(child)
else:
child = ET.Element(key)
child.text = str(val) if val is not None else "none"
elem.append(child)
return elem
def create_sequence_structure(self, video_name, total_duration):
sequence = ET.Element("sequence", id="Sequence 1")
ET.SubElement(sequence, "updatebehavior").text = "add"
ET.SubElement(sequence, "name").text = "Sequence 1"
ET.SubElement(sequence, "duration").text = str(total_duration)
rate = ET.SubElement(sequence, "rate")
ET.SubElement(rate, "ntsc").text = "TRUE"
ET.SubElement(rate, "timebase").text = "30"
timecode = ET.SubElement(sequence, "timecode")
rate_timecode = ET.SubElement(timecode, "rate")
ET.SubElement(rate_timecode, "ntsc").text = "TRUE"
ET.SubElement(rate_timecode, "timebase").text = "30"
ET.SubElement(timecode, "frame").text = "107891"
ET.SubElement(timecode, "source").text = "source"
ET.SubElement(timecode, "displayformat").text = "DF"
ET.SubElement(sequence, "in").text = "-1"
ET.SubElement(sequence, "out").text = "-1"
media = ET.SubElement(sequence, "media")
video = ET.SubElement(media, "video")
format_ = ET.SubElement(video, "format")
samplecharacteristics = ET.SubElement(format_, "samplecharacteristics")
ET.SubElement(samplecharacteristics, "width").text = "1920"
ET.SubElement(samplecharacteristics, "height").text = "1080"
ET.SubElement(samplecharacteristics, "pixelaspectratio").text = "Square"
ET.SubElement(samplecharacteristics, "anamorphic").text = "FALSE"
ET.SubElement(samplecharacteristics, "fielddominance").text = "none"
rate_sample = ET.SubElement(samplecharacteristics, "rate")
ET.SubElement(rate_sample, "ntsc").text = "TRUE"
ET.SubElement(rate_sample, "timebase").text = "30"
ET.SubElement(samplecharacteristics, "colordepth").text = "24"
track = ET.SubElement(video, "track")
return sequence, track, media
def process_video(self):
for video_path in [self.video_path]:
# Extract audio features
audio_signal = self.extract_audio(video_path)
audio_frame_features = self.process_audio(audio_signal)
# Extract video features
video = cv2.VideoCapture(video_path)
ret, frame = video.read()
if not ret:
print(f"Failed to read video: {video_path}")
continue
clip_frames = []
while len(clip_frames) < self.num_frames:
frame_processed = self.preprocess_frame(frame)
clip_frames.append(frame_processed)
ret, frame = video.read()
if not ret:
break
clip_frames = np.array(clip_frames[:self.num_frames])
video.release()
# Combine video and audio features for prediction
combined_features_for_prediction = self.combine_features(clip_frames, audio_frame_features)
# Make a prediction
prediction = self.classification_model.predict(
[np.array([combined_features_for_prediction]), np.array([audio_frame_features])])
predicted_label = self.label_encoder.inverse_transform([np.argmax(prediction)])[0]
# Retrieve clip items for the predicted label
if predicted_label in self.all_clip_items:
clip_items_for_predicted_label = self.all_clip_items[predicted_label]
print("Clip items for predicted label:", clip_items_for_predicted_label)
# Calculate the total duration
total_duration = sum([int(clip_item["duration"]) for clip_item in clip_items_for_predicted_label])
# Create the root element
root = ET.Element("xmeml", version="5")
# Create the project structure
project = ET.SubElement(root, "project")
ET.SubElement(project, "name").text = "Untitled Project 1"
children = ET.SubElement(project, "children")
# Create and append the bin element to children
bin_element = ET.Element("bin")
ET.SubElement(bin_element, "updatebehavior").text = "add"
ET.SubElement(bin_element, "name").text = "Custom Bins"
ET.SubElement(bin_element, "children") # Empty children element for the bin
children.append(bin_element)
# Define video_name_with_extension here
video_name_with_extension = os.path.basename(video_path)
# Append the clip structure directly to children (not inside bin)
clip_structure = self.create_clip_structure(video_name_with_extension, video_path,
total_duration) # <-- Added total_duration here
children.append(clip_structure)
# Append the sequence structure to children
sequence, track, media = self.create_sequence_structure(video_name_with_extension, total_duration)
children.append(sequence)
clip_item_ids = [] # List to store generated IDs
link_ids_list = [] # List to store link IDs for each clip item
video_file_elements = [] # List to store <file> elements for each video <clipitem>
for clip_item in clip_items_for_predicted_label:
# Generate three unique IDs: one for video and two for audio
video_clip_item_id = self.generate_clipitem_id(video_name_with_extension)
audio_clip_item_id_1 = self.generate_clipitem_id(video_name_with_extension)
audio_clip_item_id_2 = self.generate_clipitem_id(video_name_with_extension)
# Store the generated IDs
clip_item_ids.append(video_clip_item_id)
link_ids_list.append([video_clip_item_id, audio_clip_item_id_1, audio_clip_item_id_2])
clip_item['id'] = video_clip_item_id
clip_item['name'] = video_name_with_extension
# Convert the modified clip item dictionary to XML and append to track
clipitem_element = self.dict_to_xml("clipitem", clip_item)
# Remove all existing <link> elements from the video <clipitem>
for link in clipitem_element.findall("link"):
clipitem_element.remove(link)
# Modify the <file> section inside the <clipitem>
file_element = clipitem_element.find("file")
file_id_value = os.path.splitext(os.path.basename(video_path))[
0] # Use the video name without extension
if file_element is not None:
file_element.set("id", file_id_value) # Set the 'id' attribute
file_element.text = None # Ensure there's no text inside the <file> element
else:
# If <file> element doesn't exist, create one with the 'id' attribute
ET.SubElement(clipitem_element, "file", id=file_id_value)
# Temporarily extract the <fielddominance> element
fielddominance_elem = clipitem_element.find("fielddominance")
if fielddominance_elem is not None:
clipitem_element.remove(fielddominance_elem)
# Add the correct <linkclipref> elements
link_elem_self = ET.SubElement(clipitem_element, "link")
ET.SubElement(link_elem_self, "linkclipref").text = video_clip_item_id
link_elem_audio1 = ET.SubElement(clipitem_element, "link")
ET.SubElement(link_elem_audio1, "linkclipref").text = audio_clip_item_id_1
link_elem_audio2 = ET.SubElement(clipitem_element, "link")
ET.SubElement(link_elem_audio2, "linkclipref").text = audio_clip_item_id_2
# Re-append the <fielddominance> element to the end
if fielddominance_elem is not None:
clipitem_element.append(fielddominance_elem)
track.append(clipitem_element)
# Call the adjust_clipitem_start_end function here
self.adjust_clipitem_start_end(track.findall('clipitem'))
# Extract the <file> element and store it in the video_file_elements list
file_element = clipitem_element.find("file")
if file_element is not None:
video_file_elements.append(file_element)
# Move the audio track creation outside the clip items loop
video_clip_items_list = clip_items_for_predicted_label
# Pass the video_file_elements list to the create_audio_track function
audio_track_left = self.create_audio_track([id_list[1] for id_list in link_ids_list], -1, link_ids_list,
video_clip_items_list, video_file_elements)
audio_track_right = self.create_audio_track([id_list[2] for id_list in link_ids_list], 1, link_ids_list,
video_clip_items_list, video_file_elements)
audio = ET.SubElement(media, "audio")
audio.append(audio_track_left)
audio.append(audio_track_right)
# Save the entire XML tree to a file after processing all clip items
output_path = os.path.join(os.path.dirname(video_path), f"{os.path.basename(video_path)}_predicted.xml")
tree = ET.ElementTree(root)
with open(output_path, 'wb') as f:
tree.write(f, encoding='utf-8', xml_declaration=True)
else:
print(f"Predicted label for {video_path}: {predicted_label} (No clip items found)")
# Example usage of the class
# extractor = VideoAudioFeatureExtractor("path_to_video.mp4", "output_path.xml")
# extractor.process_video()
|