SeaLionsDetectionClassification/predict_binary_nets.py at master · HectorAnadon/SeaLionsDetectionClassification · GitHub

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
import pdb
import sys
import matplotlib.pyplot as plt
from keras.utils.io_utils import HDF5Matrix
from binary_nets import *
from global_variables import*

def predict_binary_net1(X_test, corners_test, onlyPrediction=None):
	""" Predict labels for binary net 1 and return only windows containing sealions.
	"""
	# Load training data mean
	means = np.load(PATH + 'Datasets/means_net1.npy')
	# Zero center
	X_test -= means
	# Create model
	layer, model = build_net_1(Input(shape=(X_test.shape[1], X_test.shape[2], 3)))
	# Load weights
	model.load_weights(PATH + 'Weights/weights_binary_net1.hdf5')
	# Compile model
	model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])
	# Predict model
	scores = model.predict(X_test)
	scores[:,0] = np.where(scores[:,0] > ARGMAX_THRESHOLD_1, scores[:,0], -1)
	# Get indexes of the windows labeled as sealions (0 because sealions are [1 0])
	argmax = np.argmax(scores, axis=1)
	idx = np.argwhere(argmax == 0)
	idx = np.reshape(idx, (idx.shape[0],))
	if (onlyPrediction):
		return argmax
	# Return only windows containing sealions
	return X_test[idx], corners_test[idx], scores[idx,0]


def predict_binary_net2(X_test, X_test_prev, corners_test, onlyPrediction=None):
	""" Predict labels for binary net 2 and return only windows containing sealions.
	"""
	# Load training data mean (current net)
	means = np.load(PATH + 'Datasets/means_net2.npy')
	# Zero center
	X_test -= means
	# Load training data mean (previous net)
	means = np.load(PATH + 'Datasets/means_net1.npy')
	# Zero center
	X_test_prev -= means
	# Create model
	layer, model = build_net_2(Input(shape=(X_test.shape[1], X_test.shape[2], 3)))
	# Load weights
	model.load_weights(PATH + 'Weights/weights_binary_net2.hdf5')
	# Compile model
	model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])
	# Predict model
	scores = model.predict([X_test, X_test_prev])
	scores[:,0] = np.where(scores[:,0] > ARGMAX_THRESHOLD_2, scores[:,0], -1)
	# Get indexes of the windows labeled as sealions (0 because sealions are [1 0])
	argmax = np.argmax(scores, axis=1)
	idx = np.argwhere(argmax == 0)
	idx = np.reshape(idx, (idx.shape[0],))
	if (onlyPrediction):
		return argmax
	# Return only windows containing sealions
	return X_test[idx], corners_test[idx], scores[idx,0]


def predict_binary_net3(X_test, X_test_2, X_test_1, corners_test, onlyPrediction=None):
	""" Predict labels for binary net 3 and return only windows containing sealions.
	"""
	# Load training data mean (current net)
	means = np.load(PATH + 'Datasets/means_net3.npy')
	# Zero center
	X_test -= means
	# Load training data mean (net 2)
	means = np.load(PATH + 'Datasets/means_net2.npy')
	# Zero center
	X_test_2 -= means
	# Load training data mean (net 1)
	means = np.load(PATH + 'Datasets/means_net1.npy')
	# Zero center
	X_test_1 -= means
	# Create model
	model = build_net_3(Input(shape=(X_test.shape[1], X_test.shape[2], 3)))
	# Load weights
	model.load_weights(PATH + 'Weights/weights_binary_net3.hdf5')
	# Compile model
	model.compile(loss='categorical_crossentropy', optimizer='Adam', metrics=['accuracy'])
	# Predict model
	scores = model.predict([X_test, X_test_2, X_test_1])
	scores[:,0] = np.where(scores[:,0] > ARGMAX_THRESHOLD_3, scores[:,0], -1)
	# Get indexes of the windows labeled as sealions (0 because sealions are [1 0])
	argmax = np.argmax(scores, axis=1)
	idx = np.argwhere(argmax == 0)
	idx = np.reshape(idx, (idx.shape[0],))
	if (onlyPrediction):
		return argmax
	# Return only windows containing sealions
	return X_test[idx], corners_test[idx], scores[idx,0]


"""Testing"""
if __name__ == '__main__':

	X1 = HDF5Matrix(PATH + 'Datasets/data_net1_small.h5', 'data', start=0, end=100)
	X2 = HDF5Matrix(PATH + 'Datasets/data_net2_small.h5', 'data', start=0, end=100)
	X3 = HDF5Matrix(PATH + 'Datasets/data_net3_small.h5', 'data', start=0, end=100)
	corners = np.ones((X1.shape[0], 2))
	print(X1.shape, X2.shape, X3.shape, corners.shape)

	try:
		arg1 = sys.argv[1]
	except IndexError:
		print("Command line argument missing. Usage: predict_binary_nets.py <net number>")
		sys.exit(1)

	if arg1 == '1':
		output, corners, scores = predict_binary_net1(X1, corners)
	elif arg1 == '2':
		output, corners, scores = predict_binary_net2(X2, X1, corners)
	elif arg1 == '3':
		output, corners, scores = predict_binary_net3(X3, X2, X1, corners)
	else:
		print("Wrong command line argument. Must be a value between 1-3.")

	print (output.shape)
	print (corners.shape)

	# for i in range(output.shape[0]):
	# 	img = output[i,:,:,:]
	# 	plt.imshow(img)
	# 	plt.show()