Upload 2 files
Browse files- EdAixml.py +441 -0
- usescript.py +4 -0
EdAixml.py
ADDED
@@ -0,0 +1,441 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import cv2
|
3 |
+
import pickle
|
4 |
+
import numpy as np
|
5 |
+
from keras.models import load_model
|
6 |
+
from moviepy.editor import VideoFileClip
|
7 |
+
import librosa
|
8 |
+
from sklearn.preprocessing import LabelEncoder
|
9 |
+
import xml.etree.ElementTree as ET
|
10 |
+
import random
|
11 |
+
|
12 |
+
class VideoAudioFeatureExtractor:
|
13 |
+
def __init__(self, video_path, output_path):
|
14 |
+
self.video_path = video_path
|
15 |
+
self.output_path = output_path
|
16 |
+
self.num_frames = 15
|
17 |
+
self.height = 224
|
18 |
+
self.width = 224
|
19 |
+
self.channels = 6
|
20 |
+
self.audio_feature_dim = 20
|
21 |
+
self.num_actions = 3
|
22 |
+
self.classification_model = load_model("EdAi-gamestyle.h5")
|
23 |
+
with open("label_encoder.pkl", "rb") as f:
|
24 |
+
self.label_encoder = pickle.load(f)
|
25 |
+
with open("clip_items.pkl", "rb") as f:
|
26 |
+
self.all_clip_items = pickle.load(f)
|
27 |
+
|
28 |
+
def generate_clipitem_id(self, video_name, num_digits=10):
|
29 |
+
random_number = ''.join(random.choices('0123456789', k=num_digits))
|
30 |
+
return f"{video_name}{random_number}"
|
31 |
+
|
32 |
+
def combine_features(self, video_features, audio_features):
|
33 |
+
audio_features_resized = cv2.resize(audio_features, (self.height, self.width))
|
34 |
+
audio_features_reshaped = np.repeat(audio_features_resized[:, :, np.newaxis], 3, axis=2)
|
35 |
+
audio_features_reshaped = np.repeat(audio_features_reshaped[np.newaxis, :, :, :], self.num_frames, axis=0)
|
36 |
+
combined_features = np.concatenate([video_features, audio_features_reshaped], axis=-1)
|
37 |
+
return combined_features
|
38 |
+
|
39 |
+
def extract_audio(self, video_path):
|
40 |
+
clip = VideoFileClip(video_path)
|
41 |
+
audio = clip.audio
|
42 |
+
audio_signal = audio.to_soundarray(fps=48000)
|
43 |
+
audio_signal = audio_signal.astype(np.float32)
|
44 |
+
if len(audio_signal.shape) == 2:
|
45 |
+
audio_signal = librosa.to_mono(audio_signal.T)
|
46 |
+
return audio_signal
|
47 |
+
|
48 |
+
def process_audio(self, audio_signal_segment):
|
49 |
+
mfccs = librosa.feature.mfcc(y=audio_signal_segment, sr=48000, n_mfcc=self.audio_feature_dim)
|
50 |
+
mfccs_fixed_length = librosa.util.fix_length(mfccs, size=self.num_frames)
|
51 |
+
return mfccs_fixed_length.T
|
52 |
+
|
53 |
+
def preprocess_frame(self, frame):
|
54 |
+
frame_resized = cv2.resize(frame, (self.width, self.height))
|
55 |
+
frame_normalized = (frame_resized / 255.0).astype(np.float32)
|
56 |
+
return frame_normalized
|
57 |
+
|
58 |
+
def extract_clip(self, video, audio_features, decision):
|
59 |
+
start_frame = int(decision["in"])
|
60 |
+
end_frame = int(decision["out"])
|
61 |
+
video.set(cv2.CAP_PROP_POS_FRAMES, start_frame)
|
62 |
+
clip_frames = []
|
63 |
+
for i in range(start_frame, end_frame, 2):
|
64 |
+
ret, frame = video.read()
|
65 |
+
if not ret:
|
66 |
+
break
|
67 |
+
frame_processed = self.preprocess_frame(frame)
|
68 |
+
clip_frames.append(frame_processed)
|
69 |
+
while len(clip_frames) < self.num_frames:
|
70 |
+
clip_frames.append(np.zeros((self.height, self.width, 3), dtype=np.float32))
|
71 |
+
clip_frames = np.array(clip_frames[:self.num_frames])
|
72 |
+
return clip_frames, audio_features
|
73 |
+
|
74 |
+
def create_audio_track(self, clip_item_ids, pan_value, link_ids, video_clip_items, video_file_elements):
|
75 |
+
track = ET.Element("track")
|
76 |
+
for idx, clipitem_id in enumerate(clip_item_ids):
|
77 |
+
clipitem_element = ET.Element("clipitem", id=clipitem_id)
|
78 |
+
video_clip_item = video_clip_items[idx]
|
79 |
+
in_value = video_clip_item['in']
|
80 |
+
out_value = video_clip_item['out']
|
81 |
+
start_value = video_clip_item['start']
|
82 |
+
end_value = video_clip_item['end']
|
83 |
+
ET.SubElement(clipitem_element, "name").text = video_clip_item['name']
|
84 |
+
ET.SubElement(clipitem_element, "duration").text = video_clip_item['duration']
|
85 |
+
rate_elem = ET.SubElement(clipitem_element, "rate")
|
86 |
+
ET.SubElement(rate_elem, "ntsc").text = "TRUE"
|
87 |
+
ET.SubElement(rate_elem, "timebase").text = "30"
|
88 |
+
ET.SubElement(clipitem_element, "in").text = in_value
|
89 |
+
ET.SubElement(clipitem_element, "out").text = out_value
|
90 |
+
ET.SubElement(clipitem_element, "start").text = start_value
|
91 |
+
ET.SubElement(clipitem_element, "end").text = end_value
|
92 |
+
ET.SubElement(clipitem_element, "masterclipid").text = os.path.basename(self.video_path)
|
93 |
+
sourcetrack = ET.SubElement(clipitem_element, "sourcetrack")
|
94 |
+
ET.SubElement(sourcetrack, "mediatype").text = "audio"
|
95 |
+
ET.SubElement(sourcetrack, "trackindex").text = str(idx + 1)
|
96 |
+
video_file_elem = video_clip_item.get("file")
|
97 |
+
if video_file_elem is not None:
|
98 |
+
ET.SubElement(clipitem_element, "file", id=video_file_elem.get("id"))
|
99 |
+
else:
|
100 |
+
ET.SubElement(clipitem_element, "file", id=os.path.basename(self.video_path).split('.')[0])
|
101 |
+
filter_elem = ET.SubElement(clipitem_element, "filter")
|
102 |
+
effect_elem = ET.SubElement(filter_elem, "effect")
|
103 |
+
ET.SubElement(effect_elem, "name").text = "Audio Levels"
|
104 |
+
ET.SubElement(effect_elem, "effectid").text = "audiolevels"
|
105 |
+
ET.SubElement(effect_elem, "effectcategory").text = "audiolevels"
|
106 |
+
ET.SubElement(effect_elem, "effecttype").text = "audiolevels"
|
107 |
+
ET.SubElement(effect_elem, "mediatype").text = "audio"
|
108 |
+
parameter_elem = ET.SubElement(effect_elem, "parameter")
|
109 |
+
ET.SubElement(parameter_elem, "name").text = "Level"
|
110 |
+
ET.SubElement(parameter_elem, "parameterid").text = "level"
|
111 |
+
ET.SubElement(parameter_elem, "valuemin").text = "0"
|
112 |
+
ET.SubElement(parameter_elem, "valuemax").text = "3.98109"
|
113 |
+
ET.SubElement(parameter_elem, "value").text = "1"
|
114 |
+
filter_elem = ET.SubElement(clipitem_element, "filter")
|
115 |
+
effect_elem = ET.SubElement(filter_elem, "effect")
|
116 |
+
ET.SubElement(effect_elem, "name").text = "Audio Pan"
|
117 |
+
ET.SubElement(effect_elem, "effectid").text = "audiopan"
|
118 |
+
ET.SubElement(effect_elem, "effectcategory").text = "audiopan"
|
119 |
+
ET.SubElement(effect_elem, "effecttype").text = "audiopan"
|
120 |
+
ET.SubElement(effect_elem, "mediatype").text = "audio"
|
121 |
+
parameter_elem = ET.SubElement(effect_elem, "parameter")
|
122 |
+
ET.SubElement(parameter_elem, "name").text = "Pan"
|
123 |
+
ET.SubElement(parameter_elem, "parameterid").text = "pan"
|
124 |
+
ET.SubElement(parameter_elem, "valuemin").text = "-1"
|
125 |
+
ET.SubElement(parameter_elem, "valuemax").text = "1"
|
126 |
+
ET.SubElement(parameter_elem, "value").text = str(pan_value)
|
127 |
+
for link_id in link_ids[idx]:
|
128 |
+
link_elem = ET.SubElement(clipitem_element, "link")
|
129 |
+
ET.SubElement(link_elem, "linkclipref").text = link_id
|
130 |
+
track.append(clipitem_element)
|
131 |
+
self.adjust_clipitem_start_end(track.findall('clipitem'))
|
132 |
+
return track
|
133 |
+
|
134 |
+
def create_clip_structure(self, video_name, video_path, total_duration):
|
135 |
+
clip = ET.Element("clip", id=video_name)
|
136 |
+
ET.SubElement(clip, "updatebehavior").text = "add"
|
137 |
+
ET.SubElement(clip, "name").text = video_name
|
138 |
+
ET.SubElement(clip, "duration").text = str(total_duration)
|
139 |
+
rate = ET.SubElement(clip, "rate")
|
140 |
+
ET.SubElement(rate, "ntsc").text = "TRUE"
|
141 |
+
ET.SubElement(rate, "timebase").text = "30"
|
142 |
+
ET.SubElement(clip, "in").text = "-1"
|
143 |
+
ET.SubElement(clip, "out").text = "-1"
|
144 |
+
ET.SubElement(clip, "masterclipid").text = video_name
|
145 |
+
ET.SubElement(clip, "ismasterclip").text = "TRUE"
|
146 |
+
|
147 |
+
# Logging info
|
148 |
+
logginginfo = ET.SubElement(clip, "logginginfo")
|
149 |
+
ET.SubElement(logginginfo, "scene")
|
150 |
+
ET.SubElement(logginginfo, "shottake")
|
151 |
+
ET.SubElement(logginginfo, "lognote")
|
152 |
+
ET.SubElement(logginginfo, "good").text = "FALSE"
|
153 |
+
|
154 |
+
# Labels
|
155 |
+
labels = ET.SubElement(clip, "labels")
|
156 |
+
ET.SubElement(labels, "label2")
|
157 |
+
|
158 |
+
# Comments
|
159 |
+
comments = ET.SubElement(clip, "comments")
|
160 |
+
ET.SubElement(comments, "mastercomment1")
|
161 |
+
ET.SubElement(comments, "mastercomment2")
|
162 |
+
ET.SubElement(comments, "mastercomment3")
|
163 |
+
ET.SubElement(comments, "mastercomment4")
|
164 |
+
|
165 |
+
# Media
|
166 |
+
media = ET.SubElement(clip, "media")
|
167 |
+
video = ET.SubElement(media, "video")
|
168 |
+
track = ET.SubElement(video, "track")
|
169 |
+
clipitem = ET.SubElement(track, "clipitem", id=f"{video_name}1")
|
170 |
+
ET.SubElement(clipitem, "name").text = video_name
|
171 |
+
ET.SubElement(clipitem, "duration").text = str(total_duration)
|
172 |
+
rate = ET.SubElement(clipitem, "rate")
|
173 |
+
ET.SubElement(rate, "ntsc").text = "TRUE"
|
174 |
+
ET.SubElement(rate, "timebase").text = "30"
|
175 |
+
ET.SubElement(clipitem, "in").text = "0"
|
176 |
+
ET.SubElement(clipitem, "out").text = str(total_duration)
|
177 |
+
ET.SubElement(clipitem, "start").text = "0"
|
178 |
+
ET.SubElement(clipitem, "end").text = str(total_duration)
|
179 |
+
ET.SubElement(clipitem, "pixelaspectratio").text = "Square"
|
180 |
+
ET.SubElement(clipitem, "anamorphic").text = "FALSE"
|
181 |
+
ET.SubElement(clipitem, "alphatype").text = "none"
|
182 |
+
ET.SubElement(clipitem, "masterclipid").text = video_name
|
183 |
+
|
184 |
+
# File
|
185 |
+
file_id = video_name.split('.')[0] # Extract filename without extension
|
186 |
+
file = ET.SubElement(clipitem, "file")
|
187 |
+
file.text = f'file id="{file_id}"' # Set the text of the <file> tag to the file ID without the extension
|
188 |
+
ET.SubElement(file, "name").text = video_name
|
189 |
+
ET.SubElement(file, "pathurl").text = f"file://localhost/{video_path}"
|
190 |
+
rate = ET.SubElement(file, "rate")
|
191 |
+
ET.SubElement(rate, "timebase").text = "30"
|
192 |
+
ET.SubElement(rate, "ntsc").text = "TRUE"
|
193 |
+
ET.SubElement(file, "duration").text = str(total_duration)
|
194 |
+
timecode = ET.SubElement(file, "timecode")
|
195 |
+
ET.SubElement(timecode, "string").text = "00:00:00:00"
|
196 |
+
ET.SubElement(timecode, "frame").text = "0"
|
197 |
+
ET.SubElement(timecode, "rate").text = "30"
|
198 |
+
ET.SubElement(timecode, "displayformat").text = "NDF"
|
199 |
+
media = ET.SubElement(file, "media")
|
200 |
+
video = ET.SubElement(media, "video")
|
201 |
+
audio = ET.SubElement(media, "audio")
|
202 |
+
|
203 |
+
# Audio
|
204 |
+
for i in range(2): # Assuming there are 2 audio tracks as per the example
|
205 |
+
track = ET.SubElement(audio, "track")
|
206 |
+
clipitem = ET.SubElement(track, "clipitem", id=f"{video_name}{i + 2}")
|
207 |
+
ET.SubElement(clipitem, "name").text = video_name
|
208 |
+
ET.SubElement(clipitem, "duration").text = str(total_duration)
|
209 |
+
rate = ET.SubElement(clipitem, "rate")
|
210 |
+
ET.SubElement(rate, "ntsc").text = "TRUE"
|
211 |
+
ET.SubElement(rate, "timebase").text = "30"
|
212 |
+
ET.SubElement(clipitem, "in").text = "0"
|
213 |
+
ET.SubElement(clipitem, "out").text = str(total_duration)
|
214 |
+
ET.SubElement(clipitem, "start").text = "0"
|
215 |
+
ET.SubElement(clipitem, "end").text = str(total_duration)
|
216 |
+
ET.SubElement(clipitem, "masterclipid").text = video_name
|
217 |
+
file_id = video_name.split('.')[0] # Extract filename without extension
|
218 |
+
file_elem = ET.SubElement(clipitem, "file", id=file_id)
|
219 |
+
ET.SubElement(file_elem, "name").text = video_name
|
220 |
+
ET.SubElement(file_elem, "pathurl").text = f"file://localhost/{video_path}"
|
221 |
+
sourcetrack = ET.SubElement(clipitem, "sourcetrack")
|
222 |
+
ET.SubElement(sourcetrack, "mediatype").text = "audio"
|
223 |
+
ET.SubElement(sourcetrack, "trackindex").text = str(i + 1)
|
224 |
+
|
225 |
+
return clip
|
226 |
+
|
227 |
+
def adjust_clipitem_start_end(self, clipitems):
|
228 |
+
last_end = 0
|
229 |
+
for clipitem in clipitems:
|
230 |
+
in_val = int(clipitem.find('in').text)
|
231 |
+
out_val = int(clipitem.find('out').text)
|
232 |
+
clipitem.find('start').text = str(last_end)
|
233 |
+
clipitem.find('end').text = str(last_end + (out_val - in_val))
|
234 |
+
last_end = int(clipitem.find('end').text)
|
235 |
+
|
236 |
+
def dict_to_xml(self, tag, dictionary):
|
237 |
+
attributes = dictionary.pop('attributes', {})
|
238 |
+
if tag == "clipitem" or tag == "file":
|
239 |
+
attributes["id"] = dictionary.pop("id", None)
|
240 |
+
for key, val in attributes.items():
|
241 |
+
if val is None:
|
242 |
+
attributes[key] = "none"
|
243 |
+
elem = ET.Element(tag, **attributes)
|
244 |
+
for key, val in dictionary.items():
|
245 |
+
if val is None:
|
246 |
+
val = "none"
|
247 |
+
if isinstance(val, dict):
|
248 |
+
child = self.dict_to_xml(key, val)
|
249 |
+
elem.append(child)
|
250 |
+
elif isinstance(val, list):
|
251 |
+
for item in val:
|
252 |
+
child = self.dict_to_xml(key, item)
|
253 |
+
elem.append(child)
|
254 |
+
else:
|
255 |
+
child = ET.Element(key)
|
256 |
+
child.text = str(val) if val is not None else "none"
|
257 |
+
elem.append(child)
|
258 |
+
return elem
|
259 |
+
|
260 |
+
def create_sequence_structure(self, video_name, total_duration):
|
261 |
+
sequence = ET.Element("sequence", id="Sequence 1")
|
262 |
+
ET.SubElement(sequence, "updatebehavior").text = "add"
|
263 |
+
ET.SubElement(sequence, "name").text = "Sequence 1"
|
264 |
+
ET.SubElement(sequence, "duration").text = str(total_duration)
|
265 |
+
rate = ET.SubElement(sequence, "rate")
|
266 |
+
ET.SubElement(rate, "ntsc").text = "TRUE"
|
267 |
+
ET.SubElement(rate, "timebase").text = "30"
|
268 |
+
timecode = ET.SubElement(sequence, "timecode")
|
269 |
+
rate_timecode = ET.SubElement(timecode, "rate")
|
270 |
+
ET.SubElement(rate_timecode, "ntsc").text = "TRUE"
|
271 |
+
ET.SubElement(rate_timecode, "timebase").text = "30"
|
272 |
+
ET.SubElement(timecode, "frame").text = "107891"
|
273 |
+
ET.SubElement(timecode, "source").text = "source"
|
274 |
+
ET.SubElement(timecode, "displayformat").text = "DF"
|
275 |
+
ET.SubElement(sequence, "in").text = "-1"
|
276 |
+
ET.SubElement(sequence, "out").text = "-1"
|
277 |
+
media = ET.SubElement(sequence, "media")
|
278 |
+
video = ET.SubElement(media, "video")
|
279 |
+
format_ = ET.SubElement(video, "format")
|
280 |
+
samplecharacteristics = ET.SubElement(format_, "samplecharacteristics")
|
281 |
+
ET.SubElement(samplecharacteristics, "width").text = "1920"
|
282 |
+
ET.SubElement(samplecharacteristics, "height").text = "1080"
|
283 |
+
ET.SubElement(samplecharacteristics, "pixelaspectratio").text = "Square"
|
284 |
+
ET.SubElement(samplecharacteristics, "anamorphic").text = "FALSE"
|
285 |
+
ET.SubElement(samplecharacteristics, "fielddominance").text = "none"
|
286 |
+
rate_sample = ET.SubElement(samplecharacteristics, "rate")
|
287 |
+
ET.SubElement(rate_sample, "ntsc").text = "TRUE"
|
288 |
+
ET.SubElement(rate_sample, "timebase").text = "30"
|
289 |
+
ET.SubElement(samplecharacteristics, "colordepth").text = "24"
|
290 |
+
track = ET.SubElement(video, "track")
|
291 |
+
return sequence, track, media
|
292 |
+
|
293 |
+
def process_video(self):
|
294 |
+
for video_path in [self.video_path]:
|
295 |
+
# Extract audio features
|
296 |
+
audio_signal = self.extract_audio(video_path)
|
297 |
+
audio_frame_features = self.process_audio(audio_signal)
|
298 |
+
|
299 |
+
# Extract video features
|
300 |
+
video = cv2.VideoCapture(video_path)
|
301 |
+
ret, frame = video.read()
|
302 |
+
if not ret:
|
303 |
+
print(f"Failed to read video: {video_path}")
|
304 |
+
continue
|
305 |
+
clip_frames = []
|
306 |
+
while len(clip_frames) < self.num_frames:
|
307 |
+
frame_processed = self.preprocess_frame(frame)
|
308 |
+
clip_frames.append(frame_processed)
|
309 |
+
ret, frame = video.read()
|
310 |
+
if not ret:
|
311 |
+
break
|
312 |
+
clip_frames = np.array(clip_frames[:self.num_frames])
|
313 |
+
video.release()
|
314 |
+
|
315 |
+
# Combine video and audio features for prediction
|
316 |
+
combined_features_for_prediction = self.combine_features(clip_frames, audio_frame_features)
|
317 |
+
|
318 |
+
# Make a prediction
|
319 |
+
prediction = self.classification_model.predict(
|
320 |
+
[np.array([combined_features_for_prediction]), np.array([audio_frame_features])])
|
321 |
+
predicted_label = self.label_encoder.inverse_transform([np.argmax(prediction)])[0]
|
322 |
+
|
323 |
+
# Retrieve clip items for the predicted label
|
324 |
+
if predicted_label in self.all_clip_items:
|
325 |
+
clip_items_for_predicted_label = self.all_clip_items[predicted_label]
|
326 |
+
print("Clip items for predicted label:", clip_items_for_predicted_label)
|
327 |
+
|
328 |
+
# Calculate the total duration
|
329 |
+
total_duration = sum([int(clip_item["duration"]) for clip_item in clip_items_for_predicted_label])
|
330 |
+
|
331 |
+
# Create the root element
|
332 |
+
root = ET.Element("xmeml", version="5")
|
333 |
+
|
334 |
+
# Create the project structure
|
335 |
+
project = ET.SubElement(root, "project")
|
336 |
+
ET.SubElement(project, "name").text = "Untitled Project 1"
|
337 |
+
children = ET.SubElement(project, "children")
|
338 |
+
|
339 |
+
# Create and append the bin element to children
|
340 |
+
bin_element = ET.Element("bin")
|
341 |
+
ET.SubElement(bin_element, "updatebehavior").text = "add"
|
342 |
+
ET.SubElement(bin_element, "name").text = "Custom Bins"
|
343 |
+
ET.SubElement(bin_element, "children") # Empty children element for the bin
|
344 |
+
children.append(bin_element)
|
345 |
+
|
346 |
+
# Define video_name_with_extension here
|
347 |
+
video_name_with_extension = os.path.basename(video_path)
|
348 |
+
|
349 |
+
# Append the clip structure directly to children (not inside bin)
|
350 |
+
clip_structure = self.create_clip_structure(video_name_with_extension, video_path,
|
351 |
+
total_duration) # <-- Added total_duration here
|
352 |
+
children.append(clip_structure)
|
353 |
+
|
354 |
+
# Append the sequence structure to children
|
355 |
+
sequence, track, media = self.create_sequence_structure(video_name_with_extension, total_duration)
|
356 |
+
children.append(sequence)
|
357 |
+
|
358 |
+
clip_item_ids = [] # List to store generated IDs
|
359 |
+
link_ids_list = [] # List to store link IDs for each clip item
|
360 |
+
video_file_elements = [] # List to store <file> elements for each video <clipitem>
|
361 |
+
|
362 |
+
for clip_item in clip_items_for_predicted_label:
|
363 |
+
# Generate three unique IDs: one for video and two for audio
|
364 |
+
video_clip_item_id = self.generate_clipitem_id(video_name_with_extension)
|
365 |
+
audio_clip_item_id_1 = self.generate_clipitem_id(video_name_with_extension)
|
366 |
+
audio_clip_item_id_2 = self.generate_clipitem_id(video_name_with_extension)
|
367 |
+
|
368 |
+
# Store the generated IDs
|
369 |
+
clip_item_ids.append(video_clip_item_id)
|
370 |
+
link_ids_list.append([video_clip_item_id, audio_clip_item_id_1, audio_clip_item_id_2])
|
371 |
+
|
372 |
+
clip_item['id'] = video_clip_item_id
|
373 |
+
clip_item['name'] = video_name_with_extension
|
374 |
+
|
375 |
+
# Convert the modified clip item dictionary to XML and append to track
|
376 |
+
clipitem_element = self.dict_to_xml("clipitem", clip_item)
|
377 |
+
|
378 |
+
# Remove all existing <link> elements from the video <clipitem>
|
379 |
+
for link in clipitem_element.findall("link"):
|
380 |
+
clipitem_element.remove(link)
|
381 |
+
|
382 |
+
# Modify the <file> section inside the <clipitem>
|
383 |
+
file_element = clipitem_element.find("file")
|
384 |
+
file_id_value = os.path.splitext(os.path.basename(video_path))[
|
385 |
+
0] # Use the video name without extension
|
386 |
+
if file_element is not None:
|
387 |
+
file_element.set("id", file_id_value) # Set the 'id' attribute
|
388 |
+
file_element.text = None # Ensure there's no text inside the <file> element
|
389 |
+
else:
|
390 |
+
# If <file> element doesn't exist, create one with the 'id' attribute
|
391 |
+
ET.SubElement(clipitem_element, "file", id=file_id_value)
|
392 |
+
|
393 |
+
# Temporarily extract the <fielddominance> element
|
394 |
+
fielddominance_elem = clipitem_element.find("fielddominance")
|
395 |
+
if fielddominance_elem is not None:
|
396 |
+
clipitem_element.remove(fielddominance_elem)
|
397 |
+
|
398 |
+
# Add the correct <linkclipref> elements
|
399 |
+
link_elem_self = ET.SubElement(clipitem_element, "link")
|
400 |
+
ET.SubElement(link_elem_self, "linkclipref").text = video_clip_item_id
|
401 |
+
link_elem_audio1 = ET.SubElement(clipitem_element, "link")
|
402 |
+
ET.SubElement(link_elem_audio1, "linkclipref").text = audio_clip_item_id_1
|
403 |
+
link_elem_audio2 = ET.SubElement(clipitem_element, "link")
|
404 |
+
ET.SubElement(link_elem_audio2, "linkclipref").text = audio_clip_item_id_2
|
405 |
+
|
406 |
+
# Re-append the <fielddominance> element to the end
|
407 |
+
if fielddominance_elem is not None:
|
408 |
+
clipitem_element.append(fielddominance_elem)
|
409 |
+
|
410 |
+
track.append(clipitem_element)
|
411 |
+
# Call the adjust_clipitem_start_end function here
|
412 |
+
self.adjust_clipitem_start_end(track.findall('clipitem'))
|
413 |
+
# Extract the <file> element and store it in the video_file_elements list
|
414 |
+
file_element = clipitem_element.find("file")
|
415 |
+
if file_element is not None:
|
416 |
+
video_file_elements.append(file_element)
|
417 |
+
|
418 |
+
# Move the audio track creation outside the clip items loop
|
419 |
+
video_clip_items_list = clip_items_for_predicted_label
|
420 |
+
# Pass the video_file_elements list to the create_audio_track function
|
421 |
+
audio_track_left = self.create_audio_track([id_list[1] for id_list in link_ids_list], -1, link_ids_list,
|
422 |
+
video_clip_items_list, video_file_elements)
|
423 |
+
audio_track_right = self.create_audio_track([id_list[2] for id_list in link_ids_list], 1, link_ids_list,
|
424 |
+
video_clip_items_list, video_file_elements)
|
425 |
+
|
426 |
+
audio = ET.SubElement(media, "audio")
|
427 |
+
audio.append(audio_track_left)
|
428 |
+
audio.append(audio_track_right)
|
429 |
+
|
430 |
+
# Save the entire XML tree to a file after processing all clip items
|
431 |
+
output_path = os.path.join(os.path.dirname(video_path), f"{os.path.basename(video_path)}_predicted.xml")
|
432 |
+
tree = ET.ElementTree(root)
|
433 |
+
with open(output_path, 'wb') as f:
|
434 |
+
tree.write(f, encoding='utf-8', xml_declaration=True)
|
435 |
+
else:
|
436 |
+
print(f"Predicted label for {video_path}: {predicted_label} (No clip items found)")
|
437 |
+
|
438 |
+
# Example usage of the class
|
439 |
+
# extractor = VideoAudioFeatureExtractor("path_to_video.mp4", "output_path.xml")
|
440 |
+
# extractor.process_video()
|
441 |
+
|
usescript.py
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from EdAixml import VideoAudioFeatureExtractor
|
2 |
+
|
3 |
+
extractor = VideoAudioFeatureExtractor("D:\\myscripts\\Edits\\Gaming\\Minecraft\\2ndvideo\\2023-06-21 20-21-56.mp4", "D:\\myscripts\\Edits\\Gaming\\Minecraft\\2ndvideo\\2023-06-21 20-21-56.xml")
|
4 |
+
extractor.process_video()
|