randomforest/regress.py at master · bashish101/randomforest · 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
130
131
import os
import cv2
import numpy as np
import pandas as pd
from PIL import Image
from sklearn.decomposition import PCA
from sklearn.tree import DecisionTreeRegressor

from randomforest import Forest

def impute_NaNs(dataframe):
	for key in dataframe:
		if key != 'Image':
			dataframe[key].fillna(dataframe[key].mean(), inplace = True)
	return dataframe

def hist_equalize(img, num_bins = 256):
	hist, bins = np.histogram(img.flatten(), num_bins, normed = True)
	cdf = hist.cumsum()
	cdf = (cdf / cdf[-1]) * 255.

	out_img = np.interp(img.flatten(), bins[:-1], cdf)
	out_img = out_img.reshape(img.shape)
	return out_img

def mirror_data(img_data, lab_data, img_h = 96, img_w = 96):
	out_img_data = np.copy(img_data)
	out_lab_data = np.copy(lab_data)
	out_img_data = out_img_data.reshape(-1, img_h, img_w)

	if np.random.rand() < 1:
		# Vertical flip
		out_img_data = out_img_data[:, ::-1]
		out_lab_data[:, 1::2] = img_h - out_lab_data[:, 1::2]
	else:
		# Horizontal flip
		out_img_data = out_img_data[:, :, ::-1]
		out_lab_data[:, 0::2] = img_w - out_lab_data[:, 0::2]

	out_img_data = out_img_data.reshape(-1, img_h * img_w)
	return (out_img_data, out_lab_data)

def load_data(filepath, mode = 'train', preprocess_flag = True, normalize_flag = False):
	print('Load started with mode {}'.format(mode))
	dataframe = pd.read_csv(filepath, header = 0)

	if mode == 'train':
		print('Handling NaNs...')
		if preprocess_flag:
			dataframe = impute_NaNs(dataframe)
		else:
			dataframe = dataframe.dropna()

	img_data = dataframe['Image'].apply(lambda im: np.fromstring(im, sep = ' '))
	img_data = np.vstack(img_data.values).astype(np.float32)

	if preprocess_flag:
		print('Equalizing histograms...')
		for idx in range(len(img_data)):
			img_data[idx] = hist_equalize(img_data[idx])
		if normalize_flag:
			print('Normalizing data...')
			img_data -= np.mean(img_data, axis = 0)
			img_data /= np.std(img_data, axis = 0)

	if mode == 'train':
		lab_data = dataframe.drop(['Image'], axis = 1)
		lab_data = lab_data.values.astype(np.float32)

		if preprocess_flag:
			print('Performing data augmentation...')
			img_data_aug, lab_data_aug = mirror_data(img_data, lab_data)
			img_data = np.vstack((img_data, img_data_aug))
			lab_data = np.vstack((lab_data, lab_data_aug))
	else:
		lab_data = None

	print('Load completed with mode {}'.format(mode))
	return (img_data, lab_data)

def display(img_data, lab_data, img_h = 96, img_w = 96):
	num_points = int(len(lab_data[0]) / 2)
	img_data = img_data.reshape(-1, img_h, img_w)
	color = (0, 0, 255)
	for img, lab in zip(img_data, lab_data):
		img = img.astype(np.uint8)
		test = Image.fromarray(img)
		img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
		for idx in range(num_points):
			x, y = int(lab[idx * 2]), int(lab[idx * 2 + 1])
			cv2.putText(img, str(idx), (x, y), cv2.FONT_HERSHEY_SIMPLEX, 0.25, color, 1)
		img = img[:, :, ::-1]
		img = Image.fromarray(img)
		img.show()
		input('Enter any key...')


def main():
	np.random.seed(1)

	print('Loading data...')
	train_data = load_data('./kaggle_data/training.csv', mode = 'train', preprocess_flag = False)
	test_data = load_data('./kaggle_data/test.csv', mode = 'test', preprocess_flag = False)
	X_train, Y_train = train_data[0], train_data[1]
	X_test, _ = test_data[0], test_data[1]

	inp_data = np.copy(X_test)

	print('Performing PCA...')
	pca = PCA(n_components = 23, svd_solver='randomized')
	pca.fit(X_train)

	X_train = pca.fit_transform(X_train)
	X_test = pca.fit_transform(X_test)

	print('Creating Random Forest Regressor...')

	# Uncomment to try sklearn's decision tree regressor
	# regressor = DecisionTreeRegressor(max_depth = 10)
	# regressor.fit(X_train, Y_train)

	regressor = Forest(num_trees = 1, mode = 'regression')
	regressor.fit(X_train, Y_train)

	print('Generating test predictions...')
	Y_test = regressor.predict(X_test)

	display(inp_data[-10:], Y_test[-10:])

if __name__ == '__main__':
	main()