-
Notifications
You must be signed in to change notification settings - Fork 4
/
form_predictions_for_all_tasks.py
152 lines (119 loc) · 6.46 KB
/
form_predictions_for_all_tasks.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
import os
import argparse
import pandas as pd
def combine_ftype(on_private):
# Content_2_index = {
# 0: "Empty",
# 1: "Pasta",
# 2: "Rice",
# 3: "Water"
# }
if on_private:
vggish_path = './filling_type/vggish/predictions/200903163404/ftype_private_test_agg_vggish.csv'
rf_path = './filling_type/CORSMAL-pyAudioAnalysis/ftype-randomforest-final_result_private_test.csv'
else:
vggish_path = './filling_type/vggish/predictions/200903163404/ftype_public_test_agg_vggish.csv'
rf_path = './filling_type/CORSMAL-pyAudioAnalysis/ftype-randomforest-final_result_public_test.csv'
vggish = pd.read_csv(vggish_path)
ftype_randomforest = pd.read_csv(rf_path)
ftype_randomforest = ftype_randomforest.sort_values(['Object', 'Sequence']).reset_index(drop=True)
random_forest_preds = ftype_randomforest[[
'Filling type prob0', 'Filling type prob1', 'Filling type prob2', 'Filling type prob3'
]]
vggish_preds = vggish[['ftype_prob_0', 'ftype_prob_1', 'ftype_prob_2', 'ftype_prob_3']]
ftype_combined = (random_forest_preds.values + vggish_preds.values) / 2
# return pd.Series([Content_2_index[cls] for cls in ftype_combined.argmax(axis=1)])
return pd.Series([cls for cls in ftype_combined.argmax(axis=1)])
def combine_flvl(on_private):
# filling_2_value = {0: 0, 1: 50, 2: 90}
cols_with_probs_1 = ['flvl_prob_0', 'flvl_prob_1', 'flvl_prob_2']
if on_private:
vggish_path = './filling_level/vggish/predictions/200903162117/flvl_private_test_agg_vggish.csv'
r21d_path = './filling_level/r21d_rgb/predictions/200903214601/flvl_private_test_agg_r21d_rgb.csv'
rf_path = './filling_level/CORSMAL-pyAudioAnalysis/flevel-randomforest-final_result_private_test.csv'
else:
vggish_path = './filling_level/vggish/predictions/200903162117/flvl_public_test_agg_vggish.csv'
r21d_path = './filling_level/r21d_rgb/predictions/200903214601/flvl_public_test_agg_r21d_rgb.csv'
rf_path = './filling_level/CORSMAL-pyAudioAnalysis/flevel-randomforest-final_result_public_test.csv'
flvl_vggish = pd.read_csv(vggish_path)
flvl_r21d = pd.read_csv(r21d_path)
flvl_vggish = flvl_vggish[cols_with_probs_1]
flvl_r21d = flvl_r21d[cols_with_probs_1]
# flvl_combined = (flvl_vggish.values + flvl_r21d.values) / 2
# flvl_combined = flvl_vggish.values
# we also observed that adding pyAudioAnalysis' random forest predictions, improves valid performance
cols_with_probs_2 = ['Filling level [%] prob0', 'Filling level [%] prob1', 'Filling level [%] prob2']
flvl_rf = pd.read_csv(rf_path)
flvl_rf = flvl_rf.sort_values(['Object', 'Sequence']).reset_index(drop=True)
flvl_rf = flvl_rf[cols_with_probs_2]
flvl_combined = (flvl_vggish.values + flvl_r21d.values + flvl_rf.values) / 3
# return pd.Series([int(filling_2_value[cls]) for cls in flvl_combined.argmax(axis=1)])
return pd.Series([int(cls) for cls in flvl_combined.argmax(axis=1)])
def capacity(on_private):
if on_private:
cap_path = './capacity/results/estimation_combination_private_test.csv'
# cap_path = './capacity/results/estimation_combination_with_0_private_test.csv'
# cap_path = './capacity/results/estimation_combination_with_1_private_test.csv'
else:
cap_path = './capacity/results/estimation_combination_public_test.csv'
# cap_path = './capacity/results/estimation_combination_with_0_public_test.csv'
# cap_path = './capacity/results/estimation_combination_with_1_public_test.csv'
a = pd.read_csv(cap_path)
return a['capacity[mL]']
# def estimate_fmass(submission):
# Content_2_density = {
# "Empty": 0.0, # "Empty"
# 0: 0.0, # "Empty"
# "Pasta": 0.41, # "Pasta"
# 1: 0.41, # "Pasta"
# "Rice": 0.85, # "Rice"
# 2: 0.85, # "Rice"
# "Water": 1.00, # "Water"
# 3: 1.00, # "Water"
# }
# fmass_col = []
# for cont, seq, capacity, c_mass, ftype, flvl, fmass in submission.values:
# fmass = Content_2_density[ftype] * flvl / 100 * capacity
# fmass_col.append(fmass)
# return pd.Series(fmass_col)
def make_submission_form(data_path, on_private):
columns = ['Container ID', 'Sequence', 'Filling type', 'Filling level', 'Container Capacity']
submission = pd.DataFrame(columns=columns)
if on_private:
container_ids = ['13', '14', '15']
else:
container_ids = ['10', '11', '12']
# creating columns for container id and sequence using filenames from audio folder – 0053_audio.wav -> 53
object_list = []
sequence_list = []
for container_id in container_ids:
path = os.path.join(data_path, container_id, 'audio')
filenames = sorted(os.listdir(path))
seq_ids = [int(fname.replace('_audio.wav', '')) for fname in filenames]
sequence_list.extend(seq_ids)
object_list.extend([container_id] * len(seq_ids))
submission['Container ID'] = object_list
submission['Sequence'] = sequence_list
return submission
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--predict_on_private', dest='predict_on_private', action='store_true', default=False)
parser.add_argument('--data_path', default='./dataset/')
args = parser.parse_args()
# Gather prediction for the public test set
submission_public = make_submission_form(args.data_path, on_private=False)
submission_public['Filling type'] = combine_ftype(on_private=False)
submission_public['Filling level'] = combine_flvl(on_private=False)
submission_public['Container Capacity'] = capacity(on_private=False)
# submission_public['Filling mass'] = estimate_fmass(submission_public)
submission_public.to_csv('./submission_public_test.csv', index=False)
print('Formed predictions in ./submission_public_test.csv')
# If specified, gather prediction for the public test set
if args.predict_on_private:
submission_private = make_submission_form(args.data_path, on_private=True)
submission_private['Filling type'] = combine_ftype(on_private=True)
submission_private['Filling level'] = combine_flvl(on_private=True)
submission_private['Container Capacity'] = capacity(on_private=True)
# submission_private['Filling mass'] = estimate_fmass(submission_private)
submission_private.to_csv('./submission_private_test.csv', index=False)
print('Formed predictions in ./submission_private_test.csv')