forked from udacity/CarND-Behavioral-Cloning-P3
-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
169 lines (136 loc) · 4.91 KB
/
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
import csv
import cv2
import numpy as np
import matplotlib.pyplot as plt
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split
import argparse
from keras.models import Sequential
from keras.layers import Flatten, Dense, Convolution2D, Lambda, Dropout, Cropping2D
from keras.regularizers import l2
import tensorflow as tf
from keras.preprocessing import image as keras_image
from keras.optimizers import Adam
def readAllData(root_paths):
image_paths = []
steer = []
for root_path in root_paths:
lines = []
with open(root_path + 'driving_log.csv') as csvfile:
reader = csv.reader(csvfile)
next(reader, None)
for line in reader:
lines.append(line)
for line in lines:
# skip low speeds
if float(line[6]) < 5.0:
continue
if 'recovery' in root_path and abs(float(line[3])) < 0.5:
continue
if 'recove123123ry4' in root_path and float(line[3]) > 0.0:
continue
if root_path == 'data/' and abs(float(line[3])) > 0.7:
continue
path = line[0]
filename = path.split('/')[-1]
# print(filename)
full_path = root_path + 'IMG/' + filename
# print(full_path)
measurement_steer = float(line[3])
image_paths.append(full_path)
steer.append(measurement_steer)
steer_correction = 0.2
if measurement_steer > 1.3 and 'recovery' in root_path:
path_left = root_path + 'IMG/' + line[1].split('/')[-1]
image_paths.append(path_left)
steer.append(measurement_steer + steer_correction)
if measurement_steer < -0.3 and 'recovery' in root_path:
path_right = root_path + 'IMG/' + line[2].split('/')[-1]
image_paths.append(path_right)
steer.append(measurement_steer - steer_correction)
print("File reads: " + str(len(steer)))
return (np.array(image_paths), np.array(steer))
def angleDistribution(angles):
num_bins = 20
avg_samples_per_bin = len(angles) / num_bins
plt.hist(angles, bins=num_bins)
plt.show()
return avg_samples_per_bin
def lowerZeroes(image_paths, steer, keep_prob=0.4):
image_path_new = []
steer_new = []
for i in range(len(steer)):
if abs(float(steer[i])) < 0.1 or steer[i] < -1.95:
# near-zero steer
if np.random.rand() < keep_prob:
image_path_new.append(image_paths[i])
steer_new.append(steer[i])
else:
image_path_new.append(image_paths[i])
steer_new.append(steer[i])
return image_path_new, steer_new
def flipImages(images, steer):
images_flipped = images[:,:,::-1,:]
steer_flipped = np.multiply(-1, steer)
return images_flipped, steer_flipped
def generator_data(image_paths, steer, batch_size=64):
X, y = ([], [])
image_paths, angles = shuffle(image_paths, steer)
while True:
for i in range(len(steer)):
img = cv2.cvtColor(cv2.imread(image_paths[i]), cv2.COLOR_BGR2RGB)
angle = steer[i]
img = preprocess(img)
X.append(img)
y.append(angle)
if len(X) == batch_size:
yield (np.array(X), np.array(y))
X, y = ([], [])
image_paths, angles = shuffle(image_paths, angles)
# flip horizontally and invert steer angle, if magnitude is > 0.33
if abs(float(angle)) > 0.5:
img = img[:,::-1,:]
angle = -1.0 * angle
X.append(img)
y.append(angle)
if len(X) == batch_size:
yield (np.array(X), np.array(y))
X, y = ([], [])
image_paths, angles = shuffle(image_paths, angles)
def preprocess(x):
yuv = cv2.cvtColor(x, cv2.COLOR_RGB2YUV)
new_img = yuv[60:140,:,:]
new_img = cv2.GaussianBlur(new_img, (3,3), 0)
new_img = cv2.resize(new_img,(200, 66), interpolation = cv2.INTER_AREA)
return new_img
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='CarND Behvioral Cloning')
parser.add_argument('-t', action='store_true')
args = parser.parse_args()
image_paths, steer = readAllData(['owndata-3/', 'owndata-recovery6/', 'owndata-recovery5/', 'owndata-recovery6/'])
image_paths, steer = lowerZeroes(image_paths, steer, keep_prob=0.05)
#plt.hist(steer)
#plt.show()
if args.t:
train_gen = generator_data(image_paths, steer, batch_size=128)
activation_func = 'elu'
model = Sequential()
model.add(Lambda(lambda x: x / 127.5 - 1.0, input_shape=(66, 200, 3)))
model.add(Convolution2D(24, 5, 5, activation=activation_func, border_mode='valid', subsample=(2, 2)))
model.add(Convolution2D(36, 5, 5, activation=activation_func, border_mode='valid', subsample=(2, 2)))
model.add(Convolution2D(48, 5, 5, activation=activation_func, border_mode='valid', subsample=(2, 2)))
model.add(Convolution2D(64, 3, 3, border_mode='valid', activation=activation_func))
model.add(Convolution2D(64, 3, 3, border_mode='valid', activation=activation_func))
model.add(Flatten())
model.add(Dense(100))
model.add(Dense(50))
model.add(Dense(10))
model.add(Dense(1))
adam = Adam(lr=0.0001)
model.compile(loss='mse', optimizer=adam)
print(model.summary())
print("FITTING")
history = model.fit_generator(train_gen, samples_per_epoch=20000, nb_epoch=1, verbose=1)
print("SAVING MODEL")
model.save('model.h5')
print("SAVED")