-
Notifications
You must be signed in to change notification settings - Fork 66
/
hpatch.py
134 lines (121 loc) · 4.3 KB
/
hpatch.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
import cv2
import numpy as np
from glob import glob
from joblib import Parallel, delayed
import multiprocessing
import pandas as pd
import json
import os
import time
import scipy
import copy
# all types of patches
tps = ['ref','e1','e2','e3','e4','e5','h1','h2','h3','h4','h5',\
't1','t2','t3','t4','t5']
def vis_patches(seq,tp,ids):
"""Visualises a set of types and indices for a sequence"""
w = len(tp)*65
vis = np.empty((0, w))
# add the first line with the patch type names
vis_tmp = np.empty((35, 0))
for t in tp:
tp_patch = 255*np.ones((35,65))
cv2.putText(tp_patch,t,(5,25),cv2.FONT_HERSHEY_DUPLEX , 1,0,1)
vis_tmp = np.hstack((vis_tmp,tp_patch))
vis = np.vstack((vis,vis_tmp))
# add the actual patches
for idx in ids:
vis_tmp = np.empty((65, 0))
for t in tp:
vis_tmp = np.hstack((vis_tmp,get_patch(seq,t,idx)))
vis = np.vstack((vis,vis_tmp))
return vis
def get_patch(seq,t,idx):
"""Gets a patch from a sequence with type=t and id=idx"""
return getattr(seq, t)[idx]
def get_im(seq,t):#rename this as a general method
"""Gets a patch from a sequence with type=t and id=idx"""
return getattr(seq, t)
def load_splits(f_splits):
"""Loads the json encoded splits"""
with open(f_splits) as f:
splits = json.load(f)
return splits
def load_descrs(path,dist='L2',descr_type='',sep=','):
"""Loads *all* saved patch descriptors from a root folder"""
print('>> Please wait, loading the descriptor files...')
# get all folders in the descriptor root folder, except the 1st which is '.'
t = [x[0] for x in os.walk(path)][1::]
try:
len(t) == 116
except:
print("%r does not seem like a valid HPatches descriptor root folder." % (path))
seqs_l = Parallel(n_jobs=multiprocessing.cpu_count())\
(delayed(hpatch_descr)(f,descr_type,sep) for f in t)
seqs = dict((l.name, l) for l in seqs_l)
seqs['distance'] = dist
seqs['dim'] = seqs_l[0].dim
print('>> Descriptor files loaded.')
return seqs
###############
# PCA Methods #
###############
# TODO add error if no training set - cant do PCA on test. (e.g. the
# full/view/illum split)
def compute_pcapl(descr,split):
X = np.empty((0,descr['dim']))
for seq in split['train']:
X = np.vstack((X,get_im(descr[seq],'ref')))
X -= np.mean(X, axis=0)
Xcov = np.dot(X.T,X)
Xcov = (Xcov + Xcov.T) / (2 * X.shape[0]);
d, V = np.linalg.eigh(Xcov)
vv = np.sort(d)
cl = vv[int(0.6*len(vv))]
d[d<=cl]=cl
D = np.diag(1. / np.sqrt(d))
W = np.dot(np.dot(V, D), V.T)
for seq in split['test']:
print(seq)
for t in tps:
X = get_im(descr[seq],t)
X -= np.mean(X, axis=0)
X_pca = np.dot(X,W)
X_pcapl = np.sign(X_pca) * np.power(np.abs(X_pca),0.5)
norms = np.linalg.norm(X_pcapl,axis=1)
X_proj = (X_pcapl.T / norms).T
X_proj = np.nan_to_num(X_proj)
setattr(descr[seq], t, X_proj)
################################
# Patch and descriptor classes #
################################
class hpatch_descr:
"""Class for loading an HPatches descriptor result .csv file"""
itr = tps
def __init__(self,base,descr_type='',sep=','):
self.base = base
self.name = base.split(os.path.sep)[-1]
for t in self.itr:
descr_path = os.path.join(base, t+'.csv')
df = pd.read_csv(descr_path,header=None,sep=sep).to_numpy()
df = df.astype(np.float32)
if descr_type=="bin_packed":
df = df.astype(np.uint8)
df = np.unpackbits(df, axis=1)
setattr(self, t, df)
self.N = df.shape[0]
self.dim = df.shape[1]
assert self.dim != 1, \
"Problem loading the .csv files. Please check the delimiter."
class hpatch_sequence:
"""Class for loading an HPatches sequence from a sequence folder"""
itr = tps
def __init__(self,base):
name = base.split(os.path.sep)
self.name = name[-1]
self.base = base
for t in self.itr:
im_path = os.path.join(base, t+'.png')
im = cv2.imread(im_path,0)
self.N = im.shape[0]/65
setattr(self, t, np.split(im, self.N))