def KMeans_A(rootdir, ft):
pos = []
imgspos = []
if 0 == cmp(ft, 'lbp'):
print "ft : LBP"
gbf = ilbpf.LBP_FEAT()
elif 0 == cmp(ft, 'gabor'):
print "ft : GABOR"
gbf = igbf.GABOR_FEAT()
elif 0 == cmp(ft, 'hog'):
print 'ft : HOG'
gbf = ihogf.HOG_FEAT()
elif 0 == cmp(ft, 'dwt'):
print 'ft : DWT'
gbf = idwtf.DWT_FEAT()
else:
print 'unknown ft'
return
fvs, imgs = gbf.gen_folder(rootdir, 5000)
if fvs is None:
print 'JPG None ', rootdir
return
pos.extend(fvs)
imgspos.extend(imgs)
samples = np.array(pos)
imgs = imgspos
com_num = np.minimum(300, samples.shape[0] - 10)
clf = PCA(com_num)
print 'before pca : ', samples.shape
samples = clf.fit_transform(samples)
print 'after pca : ', samples.shape
clf = KMeans(n_clusters=2, n_jobs=-2, verbose=0)
prds = clf.fit_predict(samples)
line0 = ""
line1 = ""
# line2 = ""
# line3 = ""
for k in range(len(prds)):
if prds[k] == 0:
line0 += imgs[k] + '\n'
elif prds[k] == 1:
line1 += imgs[k] + '\n'
# elif prds[k] == 2:
# line2 += imgs[k] + '\n'
# else:
# line3 += imgs[k] + '\n'
with open('A.txt', 'w') as f:
f.writelines(line0)
with open('B.txt', 'w') as f:
f.writelines(line1)
# with open('C.txt', 'w') as f:
# f.writelines(line2)
# with open('D.txt', 'w') as f:
# f.writelines(line3)
return
def DIST_B(rootdir, folderA, folderB, ft):
pos = []
neg = []
imgspos = []
imgsneg = []
if 0 == cmp(ft, 'gabor'):
print 'feature type: GABOR'
gbf = igbf.GABOR_FEAT()
elif 0 == cmp(ft, 'lbp'):
print 'feature type: LBP'
gbf = ilbpf.LBP_FEAT()
elif 0 == cmp(ft, 'hog'):
print 'feature type: HOG'
gbf = ihogf.HOG_FEAT()
elif 0 == cmp(ft, 'dwt'):
print 'feature type: DWT'
gbf = idwtf.DWT_FEAT()
else:
print 'unknown feature type'
return
#1--folderA
fvs, imgs = gbf.gen_folder(os.path.join(rootdir, folderA), 1000)
pos.extend(fvs)
imgspos.extend(imgs)
#2--folderB
fvs, imgs = gbf.gen_folder(os.path.join(rootdir, folderB), 1000)
neg.extend(fvs)
imgsneg.extend(imgs)
#3--match
samples = np.array(pos + neg)
imgs = imgspos + imgsneg
com_num = 300
if com_num + 10 > len(imgs):
com_num = len(imgs) - 10
clf_pca = PCA(com_num)
samples = clf_pca.fit_transform(samples)
print 'after pca : ', samples.shape
dists = calc_distance_set(samples[0:len(pos), :], samples[len(pos):, :], 8)
templist = []
for dist in dists:
for d, j in dist:
templist.append(imgsneg[j])
templist = list(set(templist))
lineA = ""
for imgs in templist:
lineA += imgs + '\n'
with open('A.txt', 'w') as f:
f.writelines(lineA)
return
def DIST_A(rootdir, posdir, posnum, negnum_p, ft):
pos = []
neg = []
pathpos = []
pathneg = []
folders = []
imgspos = []
imgsneg = []
with open('list.txt', 'r') as f:
for line in f:
line = line.strip()
folders.append(line)
if 0 == cmp(ft, 'gabor'):
print 'feature type: GABOR'
gbf = igbf.GABOR_FEAT()
elif 0 == cmp(ft, 'lbp'):
print 'feature type: LBP'
gbf = ilbpf.LBP_FEAT()
elif 0 == cmp(ft, 'hog'):
print 'feature type: HOG'
gbf = ihogf.HOG_FEAT()
elif 0 == cmp(ft, 'dwt'):
print 'feature type: DWT'
gbf = idwtf.DWT_FEAT()
else:
print 'unknown feature type'
return
for folder in folders:
fname = os.path.join(rootdir, folder)
if 0 == cmp(folder, posdir):
fvs, imgs = gbf.gen_folder(fname, posnum)
if fvs is None:
print 'pos None ', fname
continue
pos.extend(fvs)
imgspos.extend(imgs)
pathpos.extend([folder for k in range(len(fvs))])
else:
fvs, imgs = gbf.gen_folder(fname, negnum_p)
if fvs is None:
print 'neg None ', fname
continue
neg.extend(fvs)
imgsneg.extend(imgs)
pathneg.extend([folder for k in range(len(fvs))])
samples = np.array(pos + neg)
paths = pathpos + pathneg
imgs = imgspos + imgsneg
con_num = 300
if con_num >= len(imgs):
con_num = len(imgs) - 10
clf = PCA(con_num)
samples = clf.fit_transform(samples)
print 'after pca : ', samples.shape
topN = posnum * 3
dists = calc_distance_set(samples, samples, topN)
smap = {}
for k in range(samples.shape[0]):
pospath = paths[k]
if pospath not in smap:
smap[pospath] = {}
for dist, j in dists[k]:
negpath = paths[j]
if negpath not in smap[pospath]:
smap[pospath][negpath] = 99999.0
if smap[pospath][negpath] > dist:
smap[pospath][negpath] = dist
slist = sorted(smap.iteritems(), key=lambda k: k[0])
line = ""
for pospath, negpaths in slist:
line += pospath + '\n'
negpaths = sorted(negpaths.iteritems(),
key=lambda k: k[1],
reverse=False)
for negpath, cnt in negpaths:
line += ' ' + negpath + '(' + str(cnt) + ')\n'
with open('result.txt', 'w') as f:
f.writelines(line)
try:
shutil.rmtree('out/')
except Exception as e:
print 'catch exception ', e
os.mkdir('out/')
for pospath, negpaths in slist:
print 'create viewset for ', pospath
os.mkdir('out/' + pospath)
negpaths = sorted(negpaths.iteritems(),
key=lambda k: k[1],
reverse=False)
for negpath, cnt in negpaths:
dist = np.int32(cnt * 100)
copy_rename_jpgs(rootdir, negpath, 'out\\', pospath, 4, dist)
return
def LDA_A(rootdir, posdir, posnum, negnum_p, ft):
pos = []
neg = []
pathpos = []
pathneg = []
folders = []
imgspos = []
imgsneg = []
with open('list.txt', 'r') as f:
for line in f:
line = line.strip()
folders.append(line)
if 0 == cmp(ft, 'gabor'):
print 'feature type: GABOR'
gbf = igbf.GABOR_FEAT()
elif 0 == cmp(ft, 'hog'):
print 'feature type: HOG'
gbf = ihogf.HOG_FEAT()
elif 0 == cmp(ft, 'lbp'):
print 'feature type: LBP'
gbf = ilbpf.LBP_FEAT()
elif 0 == cmp(ft, 'dwt'):
print 'feature type: DWT'
gbf = idwtf.DWT_FEAT()
elif 0 == cmp(ft, 'yuv'):
print 'feature type: YUV'
gbf = iyuvf.YUV_FEAT()
else:
print 'unknown feature type'
return
for folder in folders:
fname = os.path.join(rootdir, folder)
if 0 == cmp(folder, posdir):
fvs, imgs = gbf.gen_folder(fname, posnum)
if fvs is None:
print 'pos None ', fname
continue
pos.extend(fvs)
imgspos.extend(imgs)
pathpos.extend([folder for k in range(len(fvs))])
else:
fvs, imgs = gbf.gen_folder(fname, negnum_p)
if fvs is None:
print 'neg None ', fname
continue
neg.extend(fvs)
imgsneg.extend(imgs)
pathneg.extend([folder for k in range(len(fvs))])
label0 = [0 for k in range(len(pos))]
label1 = [1 for k in range(len(neg))]
samples = np.array(pos + neg)
labels = np.array(label0 + label1)
paths = pathpos + pathneg
imgs = imgspos + imgsneg
#com_num = np.minimum(300, samples.shape[0] - 10)
clf = PCA(0.98)
samples = clf.fit_transform(samples)
print 'after pca : ', samples.shape
clf = LDA()
clf.fit(samples, labels)
cnf = clf.decision_function(samples)
X = []
for k in range(len(paths)):
X.append((paths[k], cnf[k], imgs[k]))
X = sorted(X, key=lambda a: a[1])
line = ""
lineA = "" #sometimes, the positive set is split into two parts
lineB = ""
for path, cnf, img in X:
line += str(cnf) + ' ' + path + ' ' + img + '\n'
if 0 != cmp(path, posdir):
continue
if cnf > 0:
lineA += img + '\n'
else:
lineB += img + '\n'
with open('A.txt', 'w') as f:
f.writelines(lineA)
with open('B.txt', 'w') as f:
f.writelines(lineB)
with open('result.txt', 'w') as f:
f.writelines(line)
return
def LDA_B(rootdir, folderA, folderB, folderC, ft):
pos = []
neg = []
imgspos = []
imgsneg = []
if 0 == cmp(ft, 'gabor'):
print 'feature type: GABOR'
gbf = igbf.GABOR_FEAT()
elif 0 == cmp(ft, 'hog'):
print 'feature type: HOG'
gbf = ihogf.HOG_FEAT()
elif 0 == cmp(ft, 'lbp'):
print 'feature type: LBP'
gbf = ilbpf.LBP_FEAT()
elif 0 == cmp(ft, 'dwt'):
print 'feature type: DWT'
gbf = idwtf.DWT_FEAT()
elif 0 == cmp(ft, 'yuv'):
print 'feature type: YUV'
gbf = iyuvf.YUV_FEAT()
else:
print 'unknown feature type'
return
#1--class A
fvs, imgs = gbf.gen_folder(os.path.join(rootdir, folderA), 1000)
pos.extend(fvs)
imgspos.extend(imgs)
#2--class B
fvs, imgs = gbf.gen_folder(os.path.join(rootdir, folderB), 1000)
neg.extend(fvs)
imgsneg.extend(imgs)
#3--train
label0 = [0 for k in range(len(pos))]
label1 = [1 for k in range(len(neg))]
samples = np.array(pos + neg)
labels = np.array(label0 + label1)
imgs = imgspos + imgsneg
print 'before pca : ', samples.shape
# com_numx = 300
# if com_numx + 10 > len(imgs):
# com_numx = len(imgs) - 10
clf_pca = PCA(0.95)
samples = clf_pca.fit_transform(samples)
print 'after pca : ', samples.shape
clf_lda = LDA()
clf_lda.fit(samples, labels)
#4--predict
fvs, imgs = gbf.gen_folder(os.path.join(rootdir, folderC), 100000)
samples = np.array(fvs)
samples = clf_pca.transform(samples)
cnf = clf_lda.decision_function(samples)
X = []
for k in range(len(imgs)):
X.append((cnf[k], imgs[k]))
X = sorted(X, key=lambda a: a[0])
lineA = "" #sometimes, the positive set is split into two parts
lineB = ""
for cnf, img in X:
if cnf > 0:
lineA += img + '\n'
else:
lineB += img + '\n'
with open('A.txt', 'w') as f:
f.writelines(lineA)
with open('B.txt', 'w') as f:
f.writelines(lineB)
return
def KNN_A(rootdir, posdir, posnum, negnum_p):
pos = []
neg = []
pathpos = []
pathneg = []
folders = []
imgspos = []
imgsneg = []
with open('list.txt', 'r') as f:
for line in f:
line = line.strip()
folders.append(line)
gbf = igbf.GABOR_FEAT()
for folder in folders:
fname = os.path.join(rootdir, folder)
if 0 == cmp(folder, posdir):
fvs, imgs = gbf.gen_folder(fname, posnum)
if fvs is None:
print 'pos None ', fname
continue
pos.extend(fvs)
imgspos.extend(imgs)
pathpos.extend([folder for k in range(len(fvs))])
else:
fvs, imgs = gbf.gen_folder(fname, negnum_p)
if fvs is None:
print 'neg None ', fname
continue
neg.extend(fvs)
imgsneg.extend(imgs)
pathneg.extend([folder for k in range(len(fvs))])
label0 = [0 for k in range(len(pos))]
label1 = [1 for k in range(len(neg))]
samples = np.array(pos + neg)
labels = np.array(label0 + label1)
paths = pathpos + pathneg
imgs = imgspos + imgsneg
clf = PCA(100)
print 'before pca : ', samples.shape
samples = clf.fit_transform(samples)
print 'after pca : ', samples.shape
if 0:
clf = KNeighborsClassifier(5)
clf.fit(samples, labels)
res = []
for k in range(samples.shape[0]):
prd = clf.predict(samples[k, :])
res.append((paths[k], prd))
res = sorted(res, key=lambda k: k[0])
line = ""
for path, prd in res:
line += path + ' ' + str(prd) + '\n'
with open('result.txt', 'w') as f:
f.writelines(line)
else:
clf = NearestNeighbors(5).fit(samples)
dists, idxs = clf.kneighbors(samples, 5)
line = ""
for k in range(len(idxs)):
for j in range(len(idxs[k])):
line += paths[idxs[k][j]] + ' '
line += '\n'
with open('result.txt', 'w') as f:
f.writelines(line)
return