-
Notifications
You must be signed in to change notification settings - Fork 3
/
INTEL_TAU_1080p.py
291 lines (139 loc) · 7.06 KB
/
INTEL_TAU_1080p.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Mon Dec 16 18:08:13 2019
@author: laakom
"""
import glob
from scipy import io as sio
import numpy as np
from tqdm import tqdm
#import tensorflow as tf
import os
class INTEL_TAU_DATASET:
def __init__(self, camera_names, set_names,root_path):
self.camera_names=camera_names
self.set_names=set_names
self.root_path = root_path
self.train_cam = None
self.splits = None
def get_train_val_test_percamera_ids( self,train_cam = 'Canon_5DSR'):
"Return a dicionary of the (train,validation,test) split and their ground truth"
self.train_cam = train_cam
partition = {}
ground_truths = {}
print('generating data for training with camera: ', train_cam)
image_path = self.root_path + train_cam + '//'
[train_ids, train_ground_truths]= self.generate_data(image_path)
assert(len(train_ids) == len(train_ground_truths))
partition['train'] = train_ids
ground_truths['train'] = train_ground_truths
#get the validation camera name
val_cam_index = (self.camera_names.index(self.train_cam)+1) % 3
val_cam = self.camera_names[val_cam_index]
image_path = self.root_path + val_cam + '//'
#generate the validation images and the ground truth
[val_ids, val_ground_truths]=self.generate_data(image_path)
assert(len(val_ids) == len(val_ground_truths))
partition['validation'] = val_ids
ground_truths['validation'] = val_ground_truths
#get the test camera name
test_cam_index = (self.camera_names.index(self.train_cam)+2) % 3
test_cam = self.camera_names[test_cam_index]
#create the original images path
image_path = self.root_path + test_cam + '//'
#generate the test images ids and the ground truth dictionary
[test_ids, test_ground_truths]=self.generate_data(image_path)
assert(len(test_ids) == len(test_ground_truths))
partition['test'] = test_ids
ground_truths['test'] = test_ground_truths
return partition,ground_truths
def generate_data(self,image_path, set_names=['field_3_cameras','field_1_cameras','lab_printouts','lab_realscene'] ):
ids = []
ground_truths = {}
for current_set in set_names:
files = glob.glob( image_path + current_set + '//' + '*.tiff')
for file in files:
ids.append(os.path.realpath(file))
ground_truths[os.path.realpath(file)]= self.get_ground_truth(file)
return ids,ground_truths
def get_ground_truth(self,file):
'read the ground truth file corresponding to an image file'
groundtruthfile = file[:-len('.tiff')]
#we use the same format of the subset files and the ground truth files:
index = groundtruthfile.rfind('_')
groundtruthfile = groundtruthfile[:index+1] + str( int(groundtruthfile[index+1:])).zfill(3) + '.wp'
gt =[]
with open(groundtruthfile, "r") as f:
groundtruth = f.read().split()
for x in groundtruth:
gt.append(float(x))
#the ground truth is saved in [R,G,B ]form
ground_truth = np.asarray(gt)
return ground_truth/ np.linalg.norm(ground_truth,2)
def read_ccm(self,camera_1,camera_2,root_path):
#camera_1 is the reference camera
#camera_2 is the source camera
#reading the mat files containing the CCM matrices
ccm_folder = root_path + 'info/Info//'
ccm = sio.loadmat(ccm_folder + 'CCMmatrices.mat')
ccm = ccm['CCMat']
#getting the indices of the CCM matrice cell, e.g., if want to get the CCM
#to trasform nikon to canon that is saved in ccm[1][0]: 1 is the index of nikon
#and 0 is the index of canon in 'camera_names' set.
indx1 = self.camera_names.index(camera_1)
indx2 = self.camera_names.index(camera_2)
return np.float64(ccm[indx2,indx1])
def set_subsets_splits(self) :
import os
dir = self.root_path + '/splits'
splits = {}
for i in tqdm(range(10)):
file = dir + '//Intel-TAU_v3_filelist_subset' + str(i + 1).zfill(2) + '.txt'
gt = []
with open(file, "r") as f:
curentfile = f.read().split()
for x in curentfile:
temp = x[ len('y:\Intel-TAU_v3'): - len('.plain16')] + '.tiff'
gt.append(self.root_path+ temp.replace('\\' , os.sep))
splits[str(i + 1).zfill(2)] = gt
self.splits = splits
def get_groundtruthfold(self,images) :
ids = []
ground_truths = {}
for current_image in images:
#we use the same format of the subset files and the ground truth files:
image_file =current_image[:-len('.tiff')]
index = image_file.rfind('_')
imagefile = image_file[:index+1] + str( int(image_file[index+1:])).zfill(3) + '.tiff'
ids.append(imagefile)
ground_truths[imagefile]= self.get_ground_truth(current_image)
return ids,ground_truths
def get_train__test_10folds( self,fold =0):
"Return a dicionary of the (train,validation,test) split and their ground truth"
#self.train_cam = train_cam
partition = {}
ground_truths = {}
partition['train'] = []
for current in range(10):
if current ==fold:
partition['test'] = self.splits[str(current + 1).zfill(2)]
else:
partition['train'].extend( self.splits[str(current + 1).zfill(2)])
from sklearn.model_selection import train_test_split
x_train, x_val = train_test_split(partition['train'], test_size=int(len(partition['train'])*0.25))
partition['train']= x_train
partition['validation']= x_val
[train_ids, train_ground_truths]= self.get_groundtruthfold( partition['train'])
assert(len(train_ids) == len(train_ground_truths))
partition['train'] = train_ids
ground_truths['train'] = train_ground_truths
[val_ids, val_ground_truths]= self.get_groundtruthfold( partition['validation'])
assert(len(train_ids) == len(train_ground_truths))
partition['validation'] = val_ids
ground_truths['validation'] = val_ground_truths
[test_ids, test_ground_truths]=self.get_groundtruthfold( partition['test'])
assert(len(test_ids) == len(test_ground_truths))
partition['test'] = test_ids
ground_truths['test'] = test_ground_truths
return partition,ground_truths