-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathentropy_model.py
176 lines (150 loc) · 6.55 KB
/
entropy_model.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
"""
Trains a model for entropy estimation of 3D objects.
Parses prevoxelization.py output as x-data and generate_entropy_dataset.py
as y-data to train a model that estimates 60 entropy values from (56,56,56)
occupancy voxel grids.
"""
import os
import sys
import argparse
from datetime import datetime
import numpy as np
import pandas as pd
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers
import utility
import kerastuner as kt
from kerastuner.tuners import Hyperband
from tqdm import tqdm
print(f"Tensorflow v{tf.__version__}\n")
parser = argparse.ArgumentParser()
parser.add_argument('--voxel_data', required=True)
parser.add_argument('--entropy_dataset', required=True)
parser.add_argument('-b', '--batch_size', type=int, default=8)
parser.add_argument('-e', '--epochs', type=int, default=5)
parser.add_argument('--load_model')
parser.add_argument('--out', default="./")
args = parser.parse_args()
TIMESTAMP = datetime.now().strftime('%d-%m-%H%M')
BASE_DIR = sys.path[0]
DATA_DIR = os.path.join(BASE_DIR, args.voxel_data)
MODEL_DIR = os.path.join(BASE_DIR, args.out, f"entropy-model")
METRICS = [
keras.metrics.AUC(name='auc'),
keras.metrics.MeanSquaredError(name='mse')
]
CALLBACKS = [
# tf.keras.callbacks.EarlyStopping(patience=3),
tf.keras.callbacks.ModelCheckpoint(
filepath=os.path.join(MODEL_DIR, 'entropy_model.h5'),
monitor='val_loss',
mode='min',
save_best_only=True,
save_freq='epoch'),
tf.keras.callbacks.TensorBoard(log_dir=os.path.join(MODEL_DIR, 'logs/')),
tf.keras.callbacks.CSVLogger(os.path.join(MODEL_DIR, 'logs/training_log.csv')),
tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss',
factor=0.3,
patience=10,
verbose=1,
mode='min',
min_lr=3e-7),
]
CLASSES = utility.CLASSES
def load_data(data, csv):
x_train, y_train, x_test, y_test = [], [], [], []
csv = pd.read_csv(csv)
print('[INFO] Loading Data...')
for lab in CLASSES:
# print(f"[DEBUG] Loading {lab}\n")
for file in tqdm(os.listdir(os.path.join(data, lab, 'train'))):
if '.npy' in file:
_data = np.load(os.path.join(data, lab, 'train', file))
padded_data = np.pad(_data, 3, 'constant')
x_train.append(padded_data)
filename = file.split(".")[0]
index = int(filename.split("_")[-1])
# print(f"[DEBUG] label, index : {lab}, {index}")
subcsv = csv[csv['label'] == lab]
entropies = np.array(subcsv[subcsv['object_index'] == index].sort_values(by=['view_code']).entropy)
# print(f"[DEBUG] Entropies of {file} : {entropies}")
y_train.append(entropies)
for file in tqdm(os.listdir(os.path.join(data, lab, 'test'))):
if '.npy' in file:
_data = np.load(os.path.join(data, lab, 'test', file))
padded_data = np.pad(_data, 3, 'constant')
x_test.append(padded_data)
filename = file.split(".")[0]
index = int(filename.split("_")[-1])
# print(f"[DEBUG] label, index : {lab}, {index}")
subcsv = csv[csv['label'] == lab]
entropies = np.array(subcsv[subcsv['object_index'] == index].sort_values(by=['view_code']).entropy)
# print(f"[DEBUG] Entropies of {file} : {entropies}")
y_test.append(entropies)
x_train = np.array(x_train)
y_train = np.array(y_train)
x_test = np.array(x_test)
y_test = np.array(y_test)
xy_train = list(zip(x_train, y_train))
np.random.shuffle(xy_train)
x_train, y_train = zip(*xy_train)
x_train = np.array(x_train)
y_train = np.array(y_train)
return x_train, y_train, x_test, y_test
def generate_cnn():
"""
Function to generate a Convolutional Neural Network
to estimate entropy from (56,56,56) voxel occupancy grids.
:return: Keras.Model
"""
inputs = keras.Input(shape=(56, 56, 56))
base = layers.Reshape(target_shape=(56, 56, 56, 1))(inputs)
# cnn_a_filters = hp.Int('cnn1_filters', min_value=4, max_value=16, step=4)
a = layers.Conv3D(8, (5, 5, 5), activation='relu', padding='same')(base)
a = layers.AveragePooling3D(pool_size=(2, 2, 2))(a)
a = layers.BatchNormalization()(a)
a = layers.Dropout(0.25)(a)
a = layers.Flatten()(a)
# cnn_b_filters = hp.Int('cnn2_filters', min_value=4, max_value=16, step=4)
b = layers.Conv3D(8, (3, 3, 3), activation='relu', padding='same')(base)
b = layers.AveragePooling3D(pool_size=(2, 2, 2))(b)
b = layers.BatchNormalization()(b)
b = layers.Dropout(0.25)(b)
b = layers.Flatten()(b)
x = layers.Concatenate(axis=1)([a, b])
# dense_units = hp.Int('dense_units', min_value=256, max_value=512, step=64)
x = layers.Dense(512, activation='relu')(x)
x = layers.BatchNormalization()(x)
x = layers.Dropout(0.5)(x)
outputs = layers.Dense(60, activation='linear')(x)
model = keras.Model(inputs=inputs, outputs=outputs, name='entronet')
model.compile(optimizer=keras.optimizers.Adam(learning_rate=5e-5), loss='mae', metrics=['mse'])
model.summary()
return model
def main():
os.mkdir(MODEL_DIR)
x_train, y_train, x_test, y_test = load_data(args.voxel_data, args.entropy_dataset)
## Uncomment following to perform hyperparameters training.
# tuner = Hyperband(generate_cnn,
# objective=kt.Objective("val_loss", direction="min"),
# max_epochs=20,
# factor=3,
# directory='../../../../data/s3866033/fyp', # Only admits relative path, for some reason.
# project_name=f'hyperband_optimization{TIMESTAMP}')
# tuner.search(x_train, y_train, epochs=10, validation_data=(x_test, y_test))
# best_hps = tuner.get_best_hyperparameters(num_trials=1)[0]
#
# model = tuner.hypermodel.build(best_hps)
model = generate_cnn()
if args.load_model is not None:
model.load_weights(args.load_model)
print(f"[INFO] Model {args.load_model} correctly loaded.")
history = model.fit(x_train, y_train,
epochs=args.epochs,
batch_size=args.batch_size,
validation_data=(x_test, y_test),
callbacks=CALLBACKS,
shuffle=True)
if __name__ == '__main__':
main()