-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathevaluation.py
324 lines (273 loc) · 13 KB
/
evaluation.py
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
import os
import pandas as pd
def process_label_file(label_file_path):
dic_subject = {1: '15', 2: '16', 3: '19', 4: '20', 5: '21', 6: '22', 7: '23', 8: '24',
9: '25', 10: '26', 11: '27', 12: '29', 13: '30', 14: '31', 15: '32', 16: '33',
17: '34', 18: '35', 19: '36', 20: '37', 21: '38', 22: '40'}
dic_express = {'disgust1': '0101', 'disgust2': '0102', 'anger1': '0401', 'anger2': '0402', 'happy1': '0502',
'happy2': '0503', 'happy3': '0505', 'happy4': '0507', 'happy5': '0508'}
df_label = pd.DataFrame(
columns=['Video_ID', 'GT_onset', 'GT_offset', 'Duration'])
if "CAS(ME)^2code_final" in label_file_path:
df_label_original = pd.read_excel(label_file_path, header=None, engine='openpyxl')
for index, row in df_label_original.iterrows():
video_id = dic_subject[row[0]] + '_' + dic_express[row[1].split('_')[0]]
onset = row[2]
apex = row[3]
offset = row[4]
if offset == 0:
offset = apex + (apex - onset)
duration = offset - row[2]
data = dict({'Video_ID': video_id, 'GT_onset': onset, 'GT_offset': offset, 'Duration': duration})
df_label = df_label.append(data, ignore_index=True)
if "SAMM" in label_file_path:
df_label_original = pd.read_excel(label_file_path, header=9, engine='openpyxl')
for index, row in df_label_original.iterrows():
video_id = row['Filename'].split('_')[0] + '_' + row['Filename'].split('_')[1]
data = dict({'Video_ID': video_id, 'GT_onset': row['Onset'],
'GT_offset': row['Offset'], 'Duration': row['Duration']})
df_label = df_label.append(data, ignore_index=True)
return df_label
def process_result_file(df_pred, frame_len, frame_step, strategy_num,
data_type='CASME2', label_type='pred_threshold_modify'):
# strategy： if the number of 0 after 1 is lower than strategy_num, then change 0 to 1
pre_label = 0
pre_video = 0
onset_seg = 0
offset_seg = 0
onset = 0
offset = 0
count = 0
df_result = pd.DataFrame(
columns=['Video_ID', 'Predicted_onset', 'Predicted_offset', 'Duration'])
for index, row in df_pred.iterrows():
if data_type == "CASME2":
video_id = row['file_path'].split('/')[9][0:7]
segment_num = row['file_path'].split('/')[10].split('.')[0]
elif data_type == "SAMM":
video_id = row['file_path'].split('/')[8]
segment_num = row['file_path'].split('/')[9].split('.')[0]
else:
raise Exception("data_type should be either CASME2 or SAMM")
if video_id == pre_video: # the same VideoID
if pre_label == 0:
if row[label_type] == 1: # just changed from 0 to 1
onset_seg = int(segment_num)
offset_seg = int(segment_num)
pre_label = 1
elif pre_label == 1:
if row[label_type] == 0:
if count < strategy_num: # change types: 1 -> 0, 10 -> 0, 100 -> 0
count = count + 1
else: # changed type: 1000 -> 0
offset_seg = int(segment_num) - strategy_num - 1
onset = onset_seg * frame_step + 1
offset = offset_seg * frame_step + frame_len
duration = offset - onset
data = dict({'Video_ID': video_id, 'Predicted_onset': onset, 'Predicted_offset': offset,
'Duration': duration})
df_result = df_result.append(data, ignore_index=True)
# initiate all the counter
pre_label = 0
onset_seg = 0
offset_seg = 0
onset = 0
offset = 0
count = 0
else: # change types: 1 -> 1, 10 -> 1, 100 -> 1, 1000 -> 1
pre_label = 1
offset_seg = int(segment_num)
count = 0
else: # change to a new video id
# initiate all the counter
pre_video = video_id
onset_seg = 0
offset_seg = 0
onset = 0
offset = 0
pre_label = 0
count = 0
if row[label_type] == 1:
pre_label = 1
onset_seg = int(segment_num)
return df_result
def process_result_file_best_strategy(df_pred, frame_len, frame_step, strategy_folder,
data_type="CASME2", label_type="pred_threshold_modify"):
# calculate the best strategy for each subject
def find_subject_best_strategy(strategy_folder):
list = {}
for path, dir_list, file_list in os.walk(strategy_folder):
for file_name in sorted(file_list):
if not file_name.startswith('.') and (path.find("/.") == -1):
csv_file = os.path.join(path, file_name)
df = pd.read_csv(csv_file)
sorted_df = df.sort_values('F1_Score', ascending=False)
best_f1_strategy = sorted_df.iloc[0, 0]
subject = file_name.split('-')[0]
list[subject] = best_f1_strategy
return list
strategy_list = find_subject_best_strategy(strategy_folder)
print(strategy_list)
# strategy： if the number of 0 after 1 is lower than strategy_num, then change 0 to 1
pre_label = 0
pre_video = 0
onset_seg = 0
offset_seg = 0
onset = 0
offset = 0
count = 0
df_result = pd.DataFrame(
columns=['Video_ID', 'Predicted_onset', 'Predicted_offset', 'Duration'])
for index, row in df_pred.iterrows():
if data_type == "CASME2":
video_id = row['file_path'].split('/')[9][0:7]
segment_num = row['file_path'].split('/')[10].split('.')[0]
elif data_type == "SAMM":
video_id = row['file_path'].split('/')[8]
segment_num = row['file_path'].split('/')[9].split('.')[0]
else:
raise Exception("data_type should be either CASME2 or SAMM")
if video_id == pre_video: # the same VideoID
subject = video_id.split("_")[0]
strategy_num = strategy_list[subject]
if pre_label == 0:
if row[label_type] == 1: # just changed from 0 to 1
onset_seg = int(segment_num)
offset_seg = int(segment_num)
pre_label = 1
elif pre_label == 1:
if row[label_type] == 0:
if count < strategy_num: # change types: 1 -> 0, 10 -> 0, 100 -> 0
count = count + 1
else: # changed type: 1000 -> 0
offset_seg = int(segment_num) - strategy_num - 1
onset = onset_seg * frame_step + 1
offset = offset_seg * frame_step + frame_len
duration = offset - onset
data = dict({'Video_ID': video_id, 'Predicted_onset': onset, 'Predicted_offset': offset,
'Duration': duration})
df_result = df_result.append(data, ignore_index=True)
# initiate all the counter
pre_label = 0
onset_seg = 0
offset_seg = 0
onset = 0
offset = 0
count = 0
else: # change types: 1 -> 1, 10 -> 1, 100 -> 1, 1000 -> 1
pre_label = 1
offset_seg = int(segment_num)
count = 0
else: # change to a new video id
# initiate all the counter
pre_video = video_id
onset_seg = 0
offset_seg = 0
onset = 0
offset = 0
pre_label = 0
count = 0
if row[label_type] == 1:
pre_label = 1
onset_seg = int(segment_num)
return df_result
def recall_precision_f1(a_clips, m_clips, n_clips):
"""
:param a_clips: the number of true positive emotion segment
:param m_clips: the number of ground truth emotion segment
:param n_clips: the number of spotted emotion segment
:return:
"""
if m_clips != 0:
recall = float(a_clips) / float(m_clips) # A / M
else:
recall = 0
if n_clips != 0:
precision = float(a_clips) / float(n_clips) # A / N
else:
precision = 0
if (recall + precision) != 0:
f1_score = 2 * recall * precision / (recall + precision)
else:
f1_score = 0
return recall, precision, f1_score
def calc_iou(clip1, clip2):
intersection = max(0, min(clip1[1], clip2[1]) - max(clip1[0], clip2[0])) + 1
union = (clip1[1] - clip1[0] + 1) + (clip2[1] - clip2[0] + 1) - intersection
if clip1[1] <= clip1[0] or clip2[1] <= clip2[0] or union <= 0:
return 0.0
else:
return float(intersection) / float(union)
def count_pred_clips(df_label, df_result, duration_threshold, iou_threshold=0.5):
macro_df_report = micro_df_report = pd.DataFrame(
columns=['Video_ID', 'GT_onset', 'GT_offset', 'Predicted_onset', 'Predicted_offset', 'Result'])
for lbs_index, lbs_row in df_label.iterrows(): # iterate over the labels
for pred_index, pred_row in df_result.iterrows(): # iterate over the results
if pred_row['Video_ID'] == lbs_row['Video_ID']:
clip1 = [pred_row['Predicted_onset'], pred_row['Predicted_offset']]
clip2 = [lbs_row['GT_onset'], lbs_row['GT_offset']]
data = dict(
{'Video_ID': pred_row['Video_ID'], 'GT_onset': lbs_row['GT_onset'],
'GT_offset': lbs_row['GT_offset'],
'Predicted_onset': pred_row['Predicted_onset'],
'Predicted_offset': pred_row['Predicted_offset'], 'Result': 'TP'})
if lbs_row['Duration'] >= duration_threshold and pred_row['Duration'] >= duration_threshold \
and calc_iou(clip1, clip2) >= iou_threshold:
macro_df_report = macro_df_report.append(data, ignore_index=True)
elif lbs_row['Duration'] < duration_threshold and pred_row['Duration'] < duration_threshold \
and calc_iou(clip1, clip2) >= iou_threshold:
micro_df_report = micro_df_report.append(data, ignore_index=True)
return macro_df_report, micro_df_report
def evaluate_logfile(macro_log_file_path, micro_log_file_path):
# read log file and count the results
def evaluate_single_log(log_file_path):
# Initial flag and A, M, N
flag = 0
A_1 = 0
M_1 = 0
N_1 = 0
A_2 = 0
M_2 = 0
N_2 = 0
df_log = pd.read_csv(log_file_path)
for index, row in df_log.iterrows():
if row['Video_ID'] == '2.0':
flag = 1
# calculate A, M, N for CASME2
if flag == 0 and row['Result'] == 'TP':
A_1 = A_1 + 1
M_1 = M_1 + 1
N_1 = N_1 + 1
if flag == 0 and row['Result'] == 'FN':
M_1 = M_1 + 1
if flag == 0 and row['Result'] == 'FP':
N_1 = N_1 + 1
# calculate A, M, N for SAMM
if flag == 1 and row['Result'] == 'TP':
A_2 = A_2 + 1
M_2 = M_2 + 1
N_2 = N_2 + 1
if flag == 1 and row['Result'] == 'FN':
M_2 = M_2 + 1
if flag == 1 and row['Result'] == 'FP':
N_2 = N_2 + 1
return A_1, M_1, N_1, A_2, M_2, N_2
# calculate the f1-scores for MaE and ME of each dataset
MaE_A1, MaE_M1, MaE_N1, MaE_A2, MaE_M2, MaE_N2 = evaluate_single_log(macro_log_file_path)
ME_A1, ME_M1, ME_N1, ME_A2, ME_M2, ME_N2 = evaluate_single_log(micro_log_file_path)
print(MaE_A1, MaE_M1, MaE_N1)
print(ME_A1, ME_M1, ME_N1)
_, _, casme2_mae_f1 = recall_precision_f1(MaE_A1, MaE_M1, MaE_N1)
_, _, casme2_me_f1 = recall_precision_f1(ME_A1, ME_M1, ME_N1)
_, _, samm_mae_f1 = recall_precision_f1(MaE_A2, MaE_M2, MaE_N2)
_, _, samm_me_f1 = recall_precision_f1(ME_A2, ME_M2, ME_N2)
# calculate the overall f1-scores of each dataset
overall_A1 = MaE_A1 + ME_A1
overall_A2 = MaE_A2 + ME_A2
overall_M1 = MaE_M1 + ME_M1
overall_M2 = MaE_M2 + ME_M2
overall_N1 = MaE_N1 + ME_N1
overall_N2 = MaE_N2 + ME_N2
_, _, overall_casme2_f1 = recall_precision_f1(overall_A1, overall_M1, overall_N1)
_, _, overall_samm_f1 = recall_precision_f1(overall_A2, overall_M2, overall_N2)
return casme2_mae_f1, casme2_me_f1, overall_casme2_f1, samm_mae_f1, samm_me_f1, overall_samm_f1