narad commited on
Commit
5a839bf
1 Parent(s): bb70788

Upload ravdess_preprocessor.py

Browse files
Files changed (1) hide show
  1. ravdess_preprocessor.py +234 -0
ravdess_preprocessor.py ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # from pathlib import Path
2
+
3
+ # import pandas as pd
4
+ # import regex as re
5
+ # import os
6
+
7
+ # import torchaudio
8
+ # import argparse
9
+ # from tqdm import tqdm
10
+
11
+ # from collections import OrderedDict
12
+
13
+
14
+ # feat_dict = OrderedDict()
15
+ # od['Modality'] = ['full-AV', 'video-only', 'audio-only']
16
+ # od['Vocal channel'] = ['speech', 'song']
17
+ # od['Emotion'] = ['neutral', 'calm', 'happy', 'sad', 'angry', 'fearful', 'disgust', 'surprised']
18
+ # od['Emotion intensity'] = ['normal', 'strong']
19
+ # od['Statement'] = ["Kids are talking by the door", "Dogs are sitting by the door"]
20
+ # od['Repetition'] = ["1st repetition", "2nd repetition"]
21
+
22
+
23
+ # # def filename2feats(filename):
24
+ # # codes = filename.stem.split('-')
25
+ # # for i, k in enumerate(od.keys()):
26
+ # # d = {}
27
+ # # d[k] = od[k][int(codes[i])-1]
28
+ # # d['Actor'] = codes[-1]
29
+ # # d['Gender'] = 'female' if int(codes[-1]) % 2 == 0 else 'male'
30
+ # # return d
31
+
32
+ # def preprocess(data_root_path):
33
+ # output_dir = data_root_path / "RAVDESS_ser"
34
+ # for f in data_root_path.iterdir():
35
+ # print(f)
36
+ # filename2feats(filename)
37
+ # print("\n\n")
38
+
39
+
40
+
41
+ # # Filename identifiers
42
+
43
+ # # Modality (01 = full-AV, 02 = video-only, 03 = audio-only).
44
+ # # Vocal channel (01 = speech, 02 = song).
45
+ # # Emotion (01 = neutral, 02 = calm, 03 = happy, 04 = sad, 05 = angry, 06 = fearful, 07 = disgust, 08 = surprised).
46
+ # # Emotional intensity (01 = normal, 02 = strong). NOTE: There is no strong intensity for the 'neutral' emotion.
47
+ # # Statement (01 = "Kids are talking by the door", 02 = "Dogs are sitting by the door").
48
+ # # Repetition (01 = 1st repetition, 02 = 2nd repetition).
49
+ # # Actor (01 to 24. Odd numbered actors are male, even numbered actors are female).
50
+
51
+ # # Filename example: 02-01-06-01-02-01-12.mp4
52
+
53
+ # # Video-only (02)
54
+ # # Speech (01)
55
+ # # Fearful (06)
56
+ # # Normal intensity (01)
57
+ # # Statement "dogs" (02)
58
+ # # 1st Repetition (01)
59
+ # # 12th Actor (12)
60
+ # # Female, as the actor ID number is even.
61
+
62
+
63
+
64
+
65
+
66
+
67
+
68
+ # # self.data_root_path = Path(data_root_path)
69
+ # # df = pd.DataFrame()
70
+ # # for session in range(1,5):
71
+ # # print(f"Processing session {session}")
72
+ # # df = pd.concat([df, self.read_session_data(session)])
73
+
74
+ # # # Write the sliced wavs
75
+ # # print("Writing wav slices to file...")
76
+ # # sample_rate = 16000
77
+ # # for index, row in df.iterrows():
78
+ # # old_filename = str(self.data_root_path / Path(row['Path_to_Wav']))
79
+ # # new_filename = str(output_dir / (index + ".wav"))
80
+ # # waveform = self.read_audio(old_filename,
81
+ # # start=row['Time_Start'],
82
+ # # end=row['Time_End'])
83
+ # # torchaudio.save(os.path.join(new_filename),
84
+ # # src=waveform,
85
+ # # sample_rate=sample_rate)
86
+ # # df.at[index, 'Path_to_Wav'] = new_filename
87
+
88
+
89
+ # # # Write out the combined data information
90
+ # # try:
91
+ # # df.to_csv(output_filename, index=False, header=True)
92
+ # # except:
93
+ # # print("Error writing dataframe to csv.")
94
+
95
+
96
+
97
+ # # def read_session_data(self, session_id):
98
+ # # d1 = self.read_emotion_labels(session_id)
99
+ # # d2 = self.read_transcriptions(session_id)
100
+ # # return d1.join(d2)
101
+
102
+
103
+ # # def read_emotion_labels(self, session_id):
104
+ # # emo_path = Path(self.data_root_path / Path(f"Session{session_id}") / Path("dialog") / Path("EmoEvaluation"))
105
+ # # emo_files = [f for f in list(emo_path.iterdir()) if f.suffix == ".txt"]
106
+ # # df = pd.DataFrame()
107
+ # # for ef in emo_files:
108
+ # # df2 = self.read_emotion_file(ef)
109
+ # # for ri, row in df2.iterrows():
110
+ # # df2.loc[ri, 'Path_to_Wav'] = os.path.join(f"Session{session_id}",
111
+ # # "dialog", "wav",
112
+ # # row['Session_ID'] +".wav")
113
+ # # df = pd.concat([df, df2])
114
+ # # df = df.set_index('ID')
115
+ # # return df
116
+
117
+
118
+ # # def slice_audio(self, session_id):
119
+ # # for i, row in df.iterrows():
120
+ # # filename = row['Session_ID'] + ".wav"
121
+ # # wav_path = Path(self.data_root_path / Path(f"Session{session_id}") / Path("dialog") / Path("wav") / Path(filename))
122
+ # # print("wav path = ", wav_path)
123
+ # # self.read_audio(wav_path, row['Time_Start'], row['Time_End'], row['Annotations'])
124
+
125
+
126
+ # # def read_emotion_file(self, filename):
127
+ # # time_extract_pattern = "\[([0-9\.]+) - ([0-9\.]+)\] +([^ ]+) +([^ ]+) \[([^\]]+)\]"
128
+ # # df = pd.DataFrame() #columns=columns)
129
+ # # i = 0
130
+ # # with open(filename) as file:
131
+ # # lines = file.readlines()
132
+ # # lines = lines[2:] #:10]
133
+
134
+ # # while i < len(lines):
135
+ # # # Remove header
136
+ # # if match := re.search(time_extract_pattern, lines[i].replace("\t", " ")):
137
+ # # time_start = float(match.group(1))
138
+ # # time_end = float(match.group(2))
139
+ # # filename = match.group(3)
140
+ # # mys_id = match.group(4)
141
+ # # digits = [float(x) for x in match.group(5).split(", ")]
142
+ # # annotations = []
143
+ # # while lines[i] != "\n":
144
+ # # i += 1
145
+ # # if lines[i].startswith("C-"):
146
+ # # aid, anns, _ = lines[i].split("\t")
147
+ # # for an in anns.split(";")[:-1]:
148
+ # # annotations.append(an.strip())
149
+ # # elif lines[i].startswith("A-"):
150
+ # # pass
151
+
152
+ # # annotations = list(set(annotations))
153
+ # # annotations = ','.join(annotations)
154
+
155
+ # # session_id = filename[:filename.rindex("_")]
156
+ # # utt_id = filename[filename.rindex("_")+1:]
157
+
158
+ # # df2 = pd.DataFrame([{
159
+ # # 'ID': filename, # ID for join between dataframes is the filename
160
+ # # 'Session_ID': session_id,
161
+ # # 'Utterance_ID': utt_id,
162
+ # # 'Time_Start': time_start,
163
+ # # 'Time_End': time_end,
164
+ # # 'Labels': annotations}])
165
+ # # df = pd.concat([df, df2], ignore_index=True)
166
+ # # else:
167
+ # # i += 1
168
+ # # return df
169
+
170
+
171
+ # # def read_transcriptions(self, session_id):
172
+ # # df = pd.DataFrame()
173
+ # # transcripts_path = Path(self.data_root_path / Path(f"Session{session_id}") / Path("dialog") / Path("transcriptions"))
174
+ # # transcript_files = [f for f in list(transcripts_path.iterdir()) if f.suffix == ".txt"]
175
+ # # for f in transcript_files:
176
+ # # df = pd.concat([df, self.read_transcript(f)], ignore_index=True)
177
+ # # df = df.set_index('ID')
178
+ # # return df
179
+
180
+
181
+ # # def read_transcript(self, filename):
182
+ # # df = pd.DataFrame()
183
+ # # with open(filename, "r") as f:
184
+ # # for l in f.readlines():
185
+ # # cols = l.strip().split(" ")
186
+ # # if l[1] != ":" and len(cols) > 2: # There are some lines like "F:Mmhmm." that get ignored here
187
+ # # df2 = pd.DataFrame([{
188
+ # # 'ID': cols[0],
189
+ # # 'Transcription': ' '.join(cols[2:])
190
+ # # }])
191
+ # # df = pd.concat([df, df2])
192
+ # # return df
193
+
194
+
195
+ # # def read_audio(self, filename, start, end, sample_rate=16000):
196
+ # # waveform, sample_rate = torchaudio.load(filename,
197
+ # # frame_offset=int(start * sample_rate),
198
+ # # num_frames=int((end-start) * sample_rate))
199
+ # # return waveform
200
+
201
+
202
+ # # if __name__ == '__main__':
203
+ # # # osx_path = '/Users/narad/Downloads/RAVDESS_full_release'
204
+ # # # windows_path = r'C:\Users\jasonn\Desktop\ser\data\RAVDESS_full_release'
205
+
206
+ # # parser = argparse.ArgumentParser(description='Process some integers.')
207
+ # # parser.add_argument('--data_dir', type=Path, required=True,
208
+ # # help='Path to IEOMCAP release directory.')
209
+ # # parser.add_argument('--output_file', type=Path, default="data.csv",
210
+ # # help='Filename for Huggingface-compatible dataset csv file.')
211
+ # # parser.add_argument('--output_dir', type=Path, default="processed",
212
+ # # help='Directory for processed wav files')
213
+ # # args = parser.parse_args()
214
+
215
+ # # print(args)
216
+
217
+ # # reader = RAVDESS(data_root_path=args.data_dir,
218
+ # # output_filename=args.output_file)
219
+
220
+
221
+
222
+
223
+
224
+
225
+
226
+
227
+
228
+
229
+
230
+ # # columns = ['Utterance_ID',
231
+ # # 'Time_Start',
232
+ # # 'Time-End',
233
+ # # 'Annotations']
234
+