-
Notifications
You must be signed in to change notification settings - Fork 2
/
combine_scores_2019.py
325 lines (257 loc) · 11.9 KB
/
combine_scores_2019.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
from language_vectors import *
from models import *
from sklearn import preprocessing
if __name__ == '__main__':
#if ran from windows machine
#titles_path = '/medias/db3/MM_GrandChallenge/MediaMemorability/dev-set/dev-set_video-captions.txt'
# These paths are hard-coded , TODO replace by arguments
alto_captions_path='alto_captions.txt'
titles_path = '/home/portege24/Documents/devset/dev-set/dev-set_video-captions.txt'
titles= read_titles(titles_path)
alto_captions=read_alto_captions(alto_captions_path)
alto_captions_train= alto_captions[0:8000]
alto_and_titles=alto_and_titles(alto_captions_train, titles)
X_resnet=pd.read_csv('X_resnet.csv')
captions = pd.read_json('captions_list.json')
danny_captions = pd.DataFrame(data=captions)
danny_captions['captions_danny'] = captions
video_name = pd.read_json('names_list.json')
danny_captions['video'] = video_name
danny_captions['video']=danny_captions['video'].str.strip("video")
danny_embeddings = np.load('embeddings_list.npy')
alto_titles_danny = adding_danny_to_rest(alto_and_titles,danny_captions)
alto_titles_danny.to_csv('alto_titles_danny.csv')
ismail1 = pd.read_csv('results_count_log_with_deepcaptions.csv')
#ismail2 = pd.read_csv('results_bertmaxpavgpca350.csv')
#ismail3 = pd.read_csv('results_rcnn_with_deepcaptions.csv')
#ismail4 = pd.read_csv('results_selfattention_with_deepcaptions.csv')
ismail_st=ismail1['short_term_predictions']
ismail_lt = ismail1['long_term_predictions']
#ismail_st=ismail_st.astype(float)
#ismail_st = ismail_st.astype(float)
labels=pd.read_csv('/home/portege24/Documents/devset/dev-set/ground-truth/ground-truth_dev-set.csv')
labels[['video', 'poubelle']] = labels['video'].str.split(".webm", expand=True, )
labels[['poubelle2', 'video']] = labels['video'].str.split("video", expand=True, )
#df = pd.merge(labels,alto_titles_danny, on="video")
df=pd.merge(labels,alto_titles_danny,on='video')
#df=pd.concat([alto_titles_danny,labels], axis=1, join='inner')
#df.to_csv('zuco-nlp/sentiment-analysis/SST_data/titles_and_alto_captions.csv')
#df.to_csv('zuco-nlp/sentiment-analysis/SST_data/alto_captions.csv')
df_tf=obtaining_tfidf_vectors(df)
df_bow=obtaining_BOW_vectors(df)
#df_bow['video']=df['video']
#to compute Bert client vectors, if already calculated can just charge file
#Bert_client_vectors= getting_Bertclient_vectors(alto_captions_train["caption"])
#to charge Bert client vectors
#Bert_client_vectors= pd.read_csv("bert_alto.csv")
Y=labels[['short-term_memorability','long-term_memorability']].values #targets
Y_st=labels['short-term_memorability']
#print(Y_st)
Y_lt = labels['long-term_memorability']
df2=pd.read_csv('et_max_alto_titles.csv')
#df2['video'] = labels['video']
df3=pd.read_csv('et_max_titles.csv')
#df3['video'] = labels['video']
df4=pd.read_csv('eeg_max_titles_pca.csv')
#df4['video'] = labels['video']
df5=pd.read_csv('eeg_max_alto_titles_pca.csv')
df6=pd.read_csv('et_max_titles_danny_alto.csv')
df7 = pd.read_csv('eeg_max_titles_pca_danny_alto.csv')
danny_embeddings = np.load('embeddings_list.npy')
print(danny_embeddings.shape)
mean_embeddings=[]
last_column_embedding = []
for x in danny_embeddings:
mean_embeddings.append(np.mean(x, axis=0))
last_column_embedding.append(x[-1])
df_embeddings = pd.DataFrame(data=mean_embeddings)
df_embeddings['video'] =danny_captions['video']
df_emb = pd.merge(labels,df_embeddings, on='video')
#print(df_emb.columns)
df_subset=pd.DataFrame()
for i in range(1024,2047):
df_subset['{}'.format(i)]=df_emb[i]
#hmp=pd.read_csv('features.csv')
#print(df_subset)
#Y_st=df_emb['short-term_memorability']
#Y_lt = df_emb['long-term_memorability']
#print(Y_st)
df_vec_st=np.concatenate([df_subset,df7,df6], axis=1)
#print(len(df_vec_st))
df_vec_lt=np.concatenate([df_subset,df6,df7], axis=1)
#df_vec_st=hmp
#df_vec_lt=hmp
#print(len(df_vec_lt))
X_st = preprocessing.normalize(df_vec_st)
X=preprocessing.normalize(df_vec_st)
X_lt = preprocessing.normalize(df_vec_lt)
#Y=preprocessing.normalize(Y)
#X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.2,
#random_state=42) # random state for reproducability
#The predictions coming from resnet need to be checked = reran to be sure there has not been a confusion in the names
folds=[1,2,3,4]
retrain=True
two_output=True
scores_st_combi=[]
scores_lt_combi=[]
scores_st_text=[]
scores_lt_text=[]
scores_st_resnet=[]
scores_lt_resnet=[]
df_pred_text = pd.DataFrame(data=np.arange(2000), columns=['trash'])
for fold in folds:
if fold==1:
X_train = X[0:6000]
X_train_st=X_st[0:6000]
X_train_lt = X_lt[0:6000]
X_test_st=X_st[6000:8000]
X_test_lt = X_lt[6000:8000]
X_test = X[6000:8000]
Y_train_st = Y_st[0:6000]
Y_train_lt= Y_lt[0:6000]
Y_train = Y[0:6000]
Y_test_st=Y_st[6000:8000]
Y_test_lt = Y_lt[6000:8000]
Y_test=Y[6000:8000]
ismail_st_test = ismail_st[6000:8000]
ismail_lt_test = ismail_lt[6000:8000]
#res_st = pd.read_csv('po_maxlt.csv')
res_st = pd.read_csv('po_max_st_1test.csv')
#res_st['ismail']=ismail_st_test.values
res_st = res_st['0']
res_lt = pd.read_csv('p1_max_lt_1test.csv')
res_lt = res_lt['0']
column_st_name='st-1'
column_lt_name='lt-1'
elif fold==2:
# to test on first 2000
X_train_st=X_st[2000:]
X_train = X[2000:]
X_train_lt=X_lt[2000:]
X_test_st = X_st[0:2000]
X_test_lt=X_lt[0:2000]
X_test = X[0:2000]
Y_train_st=Y_st[2000:]
Y_train_lt=Y_lt[2000:]
Y_train = Y[2000:]
Y_test_st = Y_st[0:2000]
Y_test_lt=Y_lt[0:2000]
Y_test = Y[0:2000]
ismail_st_test = ismail_st[0:2000]
ismail_lt_test = ismail_lt[0:2000]
res_st = pd.read_csv('po_max_st_2test.csv')
res_st = res_st['0']
res_lt = pd.read_csv('p1_max_lt_2test.csv')
res_lt = res_lt['0']
column_st_name='st-2'
column_lt_name='lt-2'
elif fold==3:
# to test on 2nd 2000
X_train_st = np.concatenate((X_st[:2000], X_st[4000:]))
X_train = np.concatenate((X[:2000], X[4000:]))
X_train_lt = np.concatenate((X_lt[:2000], X_lt[4000:]))
Y_train_st = np.concatenate((Y_st[:2000], Y_st[4000:]))
Y_train_lt = np.concatenate((Y_lt[:2000], Y_lt[4000:]))
Y_train = np.concatenate((Y[:2000], Y[4000:]))
X_test_st = X_st[2000:4000]
X_test = X[2000:4000]
X_test_lt = X_lt[2000:4000]
Y_test_st = Y_st[2000:4000]
Y_test_lt = Y_lt[2000:4000]
Y_test=Y[2000:4000]
ismail_st_test = ismail_st[2000:4000]
ismail_lt_test = ismail_lt[2000:4000]
print(len(Y_test))
print(len(Y_train_lt))
res_st = pd.read_csv('po_max_st_3test.csv')
res_st = res_st['0']
res_lt = pd.read_csv('p1_max_lt_3test.csv')
res_lt = res_lt['0']
column_st_name='st-3'
column_lt_name='lt-3'
elif fold==4:
# np.concatenate((data_x[:4000, :], data_x[6000:, :]))
X_train_st = np.concatenate((X_st[:4000], X_st[6000:]))
X_train = np.concatenate((X[:4000], X[6000:]))
X_train_lt = np.concatenate((X_lt[:4000], X_lt[6000:]))
Y_train_st = np.concatenate((Y_st[:4000], Y_st[6000:]))
Y_train = np.concatenate((Y[:4000], Y[6000:]))
Y_train_lt = np.concatenate((Y_lt[:4000], Y_lt[6000:]))
X_test_st = X_st[4000:6000]
X_test_lt = X_lt[4000:6000]
Y_test_st = Y_st[4000:6000]
Y_test_lt = Y_lt[4000:6000]
Y_test = Y[4000:6000]
X_test = X[4000:6000]
ismail_st_test=ismail_st[4000:6000]
ismail_lt_test = ismail_st[4000:6000]
print(len(Y_test))
print(len(Y_train_lt))
res_st = pd.read_csv('po_max_st_4test.csv')
res_st = res_st['0']
res_lt = pd.read_csv('p1_max_lt_4test.csv')
res_lt = res_lt['0']
column_st_name='st-4'
column_lt_name='lt-4'
if retrain==True:
if two_output==True:
predictions=three_dense_layers_model_2_outuput(X_train, X_test, Y_train, Y_test)
predictions_st=predictions[:,0]
predictions_lt = predictions[:, 0]
else:
predictions_st=three_dense_layers_model_1_outuput(X_train_st, X_test_st, Y_train_st, Y_test_st)
predictions_lt=three_dense_layers_model_1_outuput(X_train_lt, X_test_lt, Y_train_lt, Y_test_lt)
#predictions_lt=three_dense_layers_model_1_outuput(X_train_lt, X_test_lt, Y_train_lt, Y_test_lt)
else:
df_pred_text=pd.read_csv('danny_embedding.csv')
predictions_st=df_pred_text[column_st_name]
predictions_lt = df_pred_text[column_lt_name]
#score_st_text = Get_score_ind(predictions_st,Y_test_st)
#score_lt_text = Get_score_ind(predictions_lt, Y_test_st)
#score_lt = Get_score_ind(res_text_lt, Y_test_lt)
#predictions_st
#res_st.type()
#ismail_st_test.type()
res_text_st = (0.4 * res_st) + (0.3 * np.squeeze(predictions_st)) + (0.3* ismail_st_test.values)
#res_text_st = (0.4 * ismail_st_test.values) + (0.1 * np.squeeze(predictions_st)) + (0.6 * res_st)
res_text_lt = 0.4* res_lt + 0.3* np.squeeze(predictions_st)+(0.3 * ismail_st_test.values)
score_st_combi = Get_score_ind(res_text_st, Y_test_st)
score_lt_combi = Get_score_ind(res_text_lt, Y_test_lt)
scores_st_combi.append(score_st_combi)
scores_lt_combi.append(score_lt_combi)
#scores_st_text.append(score_st_text)
df_pred_text[column_st_name] = predictions_st
df_pred_text[column_lt_name] = predictions_lt
average_st=np.mean(scores_st_combi)
print(average_st)
average_lt=np.mean(scores_lt_combi)
print(average_lt)
if retrain==True:
print('done')
#df_pred_text.to_csv('pred_text_1output_MSE.csv')
#df_pred_text.to_csv('pred_text_1output_MSE_15pca.csv')
#df_pred_text.to_csv('pred_text_1output_MSE_10_pca_abs.csv')
#df_pred_text.to_csv('pred_text_1output_MSE_10_pca_abs.csv')
#df_pred_text.to_csv('pred_text_2output_noeeg.csv')
#df_pred_text.to_csv('pred_text_2output_nodanny.csv')
#df_pred_text.to_csv('pred_text_2output.csv')
#df_pred_text.to_csv('pred_2output_absolute_error.csv')
#df_pred_text.to_csv('only_titles_bow')
#df_pred_text.to_csv('tf_ngrams_2output_allcap.csv')
#df_pred_text.to_csv('2output_eegall_et_danny_alto.csv')
#df_pred_text.to_csv('1output_eegall_et_danny_alto.csv')
#df_pred_text.to_csv('1output_eegall_et_danny_alto_200,100.csv')
#df_pred_text.to_csv('1output_eegall_et_danny_alto_100,50.csv')
#df_pred_text.to_csv('bow_et_2output.csv')
#df_pred_text.to_csv('pred_text_1output_noeeg.csv')
#df_pred_text.to_csv('pred_text_1output_noet.csv')
#df_pred_text.to_csv('Lasso_best_ST.csv')
#df_pred_text.to_csv('danny_embedding.csv')
#df_pred_text.to_csv('danny_embedding_last_frame.csv')
#df_pred_text.to_csv('hmp_2output.csv')
"""scores_st_combi=[]
scores_lt_combi=[]
scores_st_text=[]
scores_lt_text=[]
scores_st_resnet=[]
scores_lt_resnet=[]"""