-
Notifications
You must be signed in to change notification settings - Fork 1
/
__init__.py
196 lines (139 loc) · 4.96 KB
/
__init__.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
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
# -*- coding: utf-8 -*-
"""food_mnist.ipynb
Automatically generated by Colaboratory.
Original file is located at
https://colab.research.google.com/drive/1up6n5A9BhuKvEj9JWwn2CX35plsX1-U-
%%time
# clone food_mnist
! git clone https://github.com/srohit0/food_mnist.git
! ls food_mnist
"""
from __future__ import print_function
import os, sys
import cv2
import numpy as np
try:
from google.colab.patches import cv2_imshow
except Exception as e:
print ("module google.colab.patches not imported.")
def cv2_imshow(image):
cv2.imshow('food MNIST', image)
"""## labels()
---
**Returns:** dictionary of labels
"""
def labels():
lbls = {}
with open(os.path.join("food_mnist", "meta", "classes.txt")) as f:
imgTags = [x.strip() for x in f.readlines()]
lbls = {key: value for (key, value) in enumerate(imgTags)}
return lbls
"""## iLabel(lblName, lblDict)
----
**Args:**
lblName: string
lblDict: dictionary of int:string values
**Returns:**
integer enum corresponding to string label
"""
def iLabel(lblName, lblDict=None):
if lblDict is None:
lblDict = labels()
for num in lblDict.keys():
if lblDict[num] == lblName.lower():
return num;
return -1
#iLabel("bibimbap")
"""## transform_image(imgFileName, imgHeight, imgWidth)
----
**Args:**
imageFileName: string
imgHeight: integer, height of the images between as per ML model
imgWidth : integer, width of the images between as per ML model
**Returns:**
numpy array of the image with shape(imgHeight, imgWidth, 3)
**Description**
This function transforms the original image to a new dimension keeping the maximum imformation in the returned image.
1. change (portrait/landscape) orientation of imageFile if necessary
2. find crop size scale, and crop image
3. resize to width and height
"""
def transform_image(imgFileName, height, width, show=False):
image = cv2.imread(imgFileName)
imageH, imageW, _ = image.shape
if show:
print(image.shape)
cv2_imshow(image)
if ( width <=0 or height <=0 ):
return image
# Step 1: change (portrait/landscape) orientation of imageFile if necessary
if ( (width>height and imageW<imageH) or (width<height and imageW>imageH)):
rotated_image = cv2.transpose(image)
image=cv2.flip(rotated_image,flipCode=0)
imageH, imageW, _ = image.shape
if show:
print(image.shape)
cv2_imshow(image)
# Step 2: find scale and crop the image
scale = min(float(imageW)/width, float(imageH)/height)
newWidth = int(width*scale)
newHeight = int(height*scale)
if ( abs(scale-1.0) > sys.float_info.epsilon ):
image = image[
int((imageH-newHeight)/2):int((imageH+newHeight)/2),
int((imageW-newWidth)/2):int((imageW+newWidth)/2)
]
imageH, imageW, _ = image.shape
if show:
print(image.shape)
cv2_imshow(image)
# Step 3: resize to width and height
if (image.shape[0] != height or image.shape[1] != width):
image = cv2.resize(image,(width,height),interpolation=cv2.INTER_AREA)
imageH, imageW, _ = image.shape
if show:
print(image.shape)
cv2_imshow(image)
return image;
#print(transform_image("food_mnist/images/apple_pie/134.jpg", 384, 512, True).shape)
"""## load_images(imageTagFileName, imgWidth, imgHeight)
----
**Args:**
imageTagFileName: string
imgWidth : integer, width of the images between as per ML model
imgHeight: integer, height of the images between as per ML model
**Returns:**
numpy array of the image with shape(numImages, imgHeight, imgWidth, 3)
list of integers representing labels
"""
def load_images(tagFile, imgHeight, imgWidth):
lbls = labels();
all_images = []
all_labels = []
with open(tagFile) as f:
tagSet = [x.strip() for x in f.readlines()]
for tag in tagSet:
lbl, imgTag = os.path.split(tag)
iLbl = iLabel(lbl, lbls)
imgFile = os.path.join("food_mnist", "images", lbl, imgTag+".jpg" )
img = transform_image(imgFile, imgHeight, imgWidth)
all_images.append(img)
all_labels.append(iLbl)
return np.array(all_images), all_labels
"""## load_data(width, height)
**Returns:**
It returns two tuples
1. x_train, x_test: uint8 array of RGB image data with shape (num_samples, width, height, 3) from the image_data_format backend setting o either channels_first or channels_last respectively.
2. y_train, y_test: uint8 array of category labels (integers in range 0-9) with shape (num_samples,).
**Example Usage**
(x_train, y_train), (x_test, y_test) = food_mnist.load_data()
labels_dict = food_mnist.labels()
"""
def load_data(imgWidth=224, imgHeight=224):
(x_train, y_train) = load_images(os.path.join("food_mnist", "meta", "train.txt"), imgWidth, imgHeight)
(x_test, y_test) = load_images(os.path.join("food_mnist", "meta", "test.txt"), imgWidth, imgHeight)
return (x_train, y_train), (x_test, y_test)
"""%%time
if __name__ == "__main__":
(x_train, y_train), (x_test, y_test) = load_data();
"""