/
color.py
197 lines (159 loc) · 6.72 KB
/
color.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
197
# -*- coding: utf-8 -*-
from __future__ import print_function
from evaluate import distance, evaluate_class
from DB import Database
from six.moves import cPickle
import numpy as np
import scipy.misc
import itertools
import os
# configs for histogram
n_bin = 12 # histogram bins
n_slice = 3 # slice image
h_type = 'region' # global or region
d_type = 'd1' # distance type
depth = 3 # retrieved depth, set to None will count the ap for whole database
''' MMAP
depth
depthNone, region,bin12,slice3, distance=d1, MMAP 0.273745840034
depth100, region,bin12,slice3, distance=d1, MMAP 0.406007856783
depth30, region,bin12,slice3, distance=d1, MMAP 0.516738512679
depth10, region,bin12,slice3, distance=d1, MMAP 0.614047666604
depth5, region,bin12,slice3, distance=d1, MMAP 0.650125
depth3, region,bin12,slice3, distance=d1, MMAP 0.657166666667
depth1, region,bin12,slice3, distance=d1, MMAP 0.62
(exps below use depth=None)
d_type
global,bin6,d1,MMAP 0.242345913685
global,bin6,cosine,MMAP 0.184176505586
n_bin
region,bin10,slice4,d1,MMAP 0.269872790396
region,bin12,slice4,d1,MMAP 0.271520862017
region,bin6,slcie3,d1,MMAP 0.262819311357
region,bin12,slice3,d1,MMAP 0.273745840034
n_slice
region,bin12,slice2,d1,MMAP 0.266076627332
region,bin12,slice3,d1,MMAP 0.273745840034
region,bin12,slice4,d1,MMAP 0.271520862017
region,bin14,slice3,d1,MMAP 0.272386552594
region,bin14,slice5,d1,MMAP 0.266877181379
region,bin16,slice3,d1,MMAP 0.273716788003
region,bin16,slice4,d1,MMAP 0.272221031804
region,bin16,slice8,d1,MMAP 0.253823360098
h_type
region,bin4,slice2,d1,MMAP 0.23358615622
global,bin4,d1,MMAP 0.229125435746
'''
# cache dir
cache_dir = 'cache'
if not os.path.exists(cache_dir):
os.makedirs(cache_dir)
class Color(object):
def histogram(self, input, n_bin=n_bin, type=h_type, n_slice=n_slice, normalize=True):
''' count img color histogram
arguments
input : a path to a image or a numpy.ndarray
n_bin : number of bins for each channel
type : 'global' means count the histogram for whole image
'region' means count the histogram for regions in images, then concatanate all of them
n_slice : work when type equals to 'region', height & width will equally sliced into N slices
normalize: normalize output histogram
return
type == 'global'
a numpy array with size n_bin ** channel
type == 'region'
a numpy array with size n_slice * n_slice * (n_bin ** channel)
'''
if isinstance(input, np.ndarray): # examinate input type
img = input.copy()
else:
img = scipy.misc.imread(input, mode='RGB')
height, width, channel = img.shape
bins = np.linspace(0, 256, n_bin+1, endpoint=True) # slice bins equally for each channel
if type == 'global':
hist = self._count_hist(img, n_bin, bins, channel)
elif type == 'region':
hist = np.zeros((n_slice, n_slice, n_bin ** channel))
h_silce = np.around(np.linspace(0, height, n_slice+1, endpoint=True)).astype(int)
w_slice = np.around(np.linspace(0, width, n_slice+1, endpoint=True)).astype(int)
for hs in range(len(h_silce)-1):
for ws in range(len(w_slice)-1):
img_r = img[h_silce[hs]:h_silce[hs+1], w_slice[ws]:w_slice[ws+1]] # slice img to regions
hist[hs][ws] = self._count_hist(img_r, n_bin, bins, channel)
if normalize:
hist /= np.sum(hist)
return hist.flatten()
def _count_hist(self, input, n_bin, bins, channel):
img = input.copy()
bins_idx = {key: idx for idx, key in enumerate(itertools.product(np.arange(n_bin), repeat=channel))} # permutation of bins
hist = np.zeros(n_bin ** channel)
# cluster every pixels
for idx in range(len(bins)-1):
img[(input >= bins[idx]) & (input < bins[idx+1])] = idx
# add pixels into bins
height, width, _ = img.shape
for h in range(height):
for w in range(width):
b_idx = bins_idx[tuple(img[h,w])]
hist[b_idx] += 1
return hist
def make_samples(self, db, verbose=True):
if h_type == 'global':
sample_cache = "histogram_cache-{}-n_bin{}".format(h_type, n_bin)
elif h_type == 'region':
sample_cache = "histogram_cache-{}-n_bin{}-n_slice{}".format(h_type, n_bin, n_slice)
try:
samples = cPickle.load(open(os.path.join(cache_dir, sample_cache), "rb", True))
if verbose:
print("Using cache..., config=%s, distance=%s, depth=%s" % (sample_cache, d_type, depth))
except:
if verbose:
print("Counting histogram..., config=%s, distance=%s, depth=%s" % (sample_cache, d_type, depth))
samples = []
data = db.get_data()
for d in data.itertuples():
d_img, d_cls = getattr(d, "img"), getattr(d, "cls")
d_hist = self.histogram(d_img, type=h_type, n_bin=n_bin, n_slice=n_slice)
samples.append({
'img': d_img,
'cls': d_cls,
'hist': d_hist
})
cPickle.dump(samples, open(os.path.join(cache_dir, sample_cache), "wb", True))
return samples
if __name__ == "__main__":
db = Database()
data = db.get_data()
color = Color()
# test normalize
hist = color.histogram(data.ix[0,0], type='global')
assert hist.sum() - 1 < 1e-9, "normalize false"
# test histogram bins
def sigmoid(z):
a = 1.0 / (1.0 + np.exp(-1. * z))
return a
np.random.seed(0)
IMG = sigmoid(np.random.randn(2,2,3)) * 255
IMG = IMG.astype(int)
hist = color.histogram(IMG, type='global', n_bin=4)
assert np.equal(np.where(hist > 0)[0], np.array([37, 43, 58, 61])).all(), "global histogram implement failed"
hist = color.histogram(IMG, type='region', n_bin=4, n_slice=2)
assert np.equal(np.where(hist > 0)[0], np.array([58, 125, 165, 235])).all(), "region histogram implement failed"
# examinate distance
np.random.seed(1)
IMG = sigmoid(np.random.randn(4,4,3)) * 255
IMG = IMG.astype(int)
hist = color.histogram(IMG, type='region', n_bin=4, n_slice=2)
IMG2 = sigmoid(np.random.randn(4,4,3)) * 255
IMG2 = IMG2.astype(int)
hist2 = color.histogram(IMG2, type='region', n_bin=4, n_slice=2)
assert distance(hist, hist2, d_type='d1') == 2, "d1 implement failed"
assert distance(hist, hist2, d_type='d2-norm') == 2, "d2 implement failed"
# evaluate database
APs = evaluate_class(db, f_class=Color, d_type=d_type, depth=depth)
cls_MAPs = []
for cls, cls_APs in APs.items():
MAP = np.mean(cls_APs)
print("Class {}, MAP {}".format(cls, MAP))
cls_MAPs.append(MAP)
print("MMAP", np.mean(cls_MAPs))