def hog_histogram(img_list, out_file, debug):
f = open(img_list)
filenames = f.read().splitlines()
f.close()
image_vector = []
ctr = 0
for entry in filenames:
dsift.process_image_dsift(entry, 'tmp.sift', 20, 10, resize=(200, 200))
l, d = sift.read_features_from_file('tmp.sift')
if debug:
im = array(Image.open(entry))
sift.plot_features(im, l, True)
show()
image_vector.append(d.flatten().tolist())
os.remove('tmp.sift')
os.remove('tmp.frame')
if not PGM:
os.remove('tmp.pgm')
if out_file != None:
f = open(out_file, 'a')
if ctr == 0:
header = ''
for i in range(len(image_vector[ctr])):
if i == 0:
header += str(i)
else:
header += ',' + str(i)
f.write(header)
f.write('\n')
vector = ''
for i in range(len(image_vector[ctr])):
if i == 0:
vector += str(image_vector[ctr][i])
else:
vector += ',' + str(image_vector[ctr][i])
f.write(vector)
f.write('\n')
f.close()
ctr += 1
return image_vector
# -*- coding: utf-8 -*-
"""
Created on Wed May 14 15:31:48 2014
@author: admin
"""
import dsift,sift
import numpy as np
from PIL import Image
from pylab import *
dsift.process_image_dsift('empire.jpg','empire.sift',90,40,True)
l,d=sift.read_features_from_file('empire.sift')
im=np.array(Image.open('empire.jpg'))
sift.plot_features(im,l,True)
show()
####################################################################
def hog_histogram(img_list, class_label, out_file, norm, debug):
f = open(img_list)
filenames = f.read().splitlines()
f.close()
image_vector = []
ctr = 0
count = 0
end_count = len(filenames)
for entry in filenames:
#want
dsift.process_image_dsift(entry, 'tmp.sift', 20, 10, resize=(200,200))
l,d = sift.read_features_from_file('tmp.sift')
# if debug:
# cv2.destroyAllWindows()
#
# #used to draw on original image instead of resized image
# im = array(Image.open(entry))
# tmp_im = cv2.resize(im, (200,200))
# tmp_im = cv2.cvtColor(tmp_im, cv2.COLOR_BGR2GRAY)
# sift.plot_features(tmp_im, l, True)
#
# show()
image_vector.append(d.flatten().tolist())
os.remove('tmp.sift')
os.remove('tmp.frame')
if not PGM:
os.remove('tmp.pgm')
if out_file != None:
f = open(out_file, 'a')
if class_label == 1:
vector = '+1'
else:
vector = '-1'
vec_length = len(image_vector[ctr])
for i, dim in zip(xrange(vec_length), xrange(1, vec_length+1)):
vector += ' ' + str(dim) + ':' + str(image_vector[ctr][i])
f.write(vector)
f.write('\n')
f.close()
ctr += 1
count += 1
#sys.stdout.write('\r%i/%i' % (count, end_count))
#sys.stdout.flush()
return image_vector
print_error()
if not dirImage or not os.path.isdir(dirImage):
print "No/invalid image directory specified"
print_error()
if not dirFeats or not os.path.isdir(dirFeats):
print "No/invalid feature directory specified"
print_error()
imgSize=int(imgSize)
if typeFeats == 'sift':
imlist = imtools.get_imlist(dirImage)
imnbr = len(imlist)
for i in range(imnbr):
im1 = array(Image.open(imlist[i]).convert('L'))
siftName = basename(imlist[i])
sift.process_image(imlist[i], dirFeats+siftName+".sift",params="--edge-thresh 10 --peak-thresh 5", resize=(imgSize,imgSize))
#featlist.append("features/"+siftName+".sift")
if typeFeats == 'dsift':
imlist = imtools.get_imlist(dirImage)
imnbr = len(imlist)
for i in range(imnbr):
im1 = array(Image.open(imlist[i]).convert('L'))
siftName = basename(imlist[i])
dsift.process_image_dsift(imlist[i], dirFeats+siftName+".dsift",10,5,resize=(imgSize,imgSize))
#featlist.append("features/"+siftName+".dsift")
cat_paths.sort()
cats = [basename(cat_path) for cat_path in cat_paths]
return cats
def get_imgfiles(path):
all_files = []
all_files.extend([
join(path, basename(fname)) for fname in glob(path + "/*")
if splitext(fname)[-1].lower() in EXTENSIONS
])
return all_files
dataset_path = '../flower_rec1/square_images128_dsift/test'
categories = get_categories(dataset_path)
for category in categories:
category_path = join(dataset_path, category)
image_file_list = get_imgfiles(category_path)
num_images = len(image_file_list)
feature_file_list = [
image_file_list[i][:-3] + "dsift_{0}_{1}".format(size, step)
for i in range(num_images)
]
for i in range(num_images):
dsift.process_image_dsift(image_file_list[i], feature_file_list[i],
size, step)
#if img_dir != '':
# if os.path.isdir(img_dir):
# imlist = imtools.get_imlist(img_dir)
#elif ifile != '':
# if os.path.isfile(ifile):
# imlist = list(ifile)
##########################################
imlist = imtools.get_imlist('corpus/train/')
imnbr = len(imlist)
for i in range(imnbr):
im1 = array(Image.open(imlist[i]).convert('L'))
siftName = basename(imlist[i])
dsift.process_image_dsift(imlist[i], "features/train/"+siftName+".dsift",10,5,resize=(50,50))
#featlist.append("features/"+siftName+".dsift")
imlist = imtools.get_imlist('corpus/test/')
imnbr = len(imlist)
for i in range(imnbr):
im1 = array(Image.open(imlist[i]).convert('L'))
siftName = basename(imlist[i])
dsift.process_image_dsift(imlist[i], "features/test/"+siftName+".dsift",10,5,resize=(50,50))
def read_feature_labels(path):
featlist = [os.path.join(path,f) for f in os.listdir(path) if f.endswith('.dsift')]
features = []
for featfile in featlist:
l,d = sift.read_features_from_file(featfile)
def md(p):
try:
os.makedirs(p)
except:
pass
md(os.path.join(d, 'train'))
md(os.path.join(d, 'test'))
for p in train:
os.symlink(p, os.path.join(d, 'train', os.path.basename(p)))
for p in test:
os.symlink(p, os.path.join(d, 'test', os.path.basename(p)))
# Compute dsift descriptors.
imlist = glob.glob(d + '/*/*.ppm')
dsiftlist = [os.path.splitext(im)[0] + '.dsift' for im in imlist]
# About 10s:
#for im, dim in zip(imlist, dsiftlist):
# dsift.process_image_dsift(im, dim, 10, 5, resize=(50, 50))
# About 7s:
from multiprocessing.dummy import Pool as ThreadPool
pool = ThreadPool(8)
pool.map(
lambda x: dsift.process_image_dsift(x[0], x[1], 10, 5, resize=(50, 50)),
zip(imlist, dsiftlist))
pool.close()
pool.join()
#!/usr/bin/python
# -*- coding: utf-8 -*-
import os
d = 'test'
imlist = [os.path.join(d,f) for f in os.listdir(d) if f.endswith('.ppm')]
d = 'train'
imlist += [os.path.join(d,f) for f in os.listdir(d) if f.endswith('.ppm')]
print imlist
import dsift
# 画像を固定サイズ（50×50）で処理する
for filename in imlist:
featfile = filename[:-3]+'dsift'
dsift.process_image_dsift(filename,featfile,10,5,resize=(50,50))
if not dirFeats or not os.path.isdir(dirFeats):
print "No/invalid feature directory specified"
print_error()
imgSize = int(imgSize)
if typeFeats == 'sift':
imlist = imtools.get_imlist(dirImage)
imnbr = len(imlist)
for i in range(imnbr):
im1 = array(Image.open(imlist[i]).convert('L'))
siftName = basename(imlist[i])
sift.process_image(imlist[i],
dirFeats + siftName + ".sift",
params="--edge-thresh 10 --peak-thresh 5",
resize=(imgSize, imgSize))
#featlist.append("features/"+siftName+".sift")
if typeFeats == 'dsift':
imlist = imtools.get_imlist(dirImage)
imnbr = len(imlist)
for i in range(imnbr):
im1 = array(Image.open(imlist[i]).convert('L'))
siftName = basename(imlist[i])
dsift.process_image_dsift(imlist[i],
dirFeats + siftName + ".dsift",
10,
5,
resize=(imgSize, imgSize))
#featlist.append("features/"+siftName+".dsift")
def hog_histogram(img_list, class_label, out_file, norm, debug):
f = open(img_list)
filenames = f.read().splitlines()
f.close()
image_vector = []
ctr = 0
count = 0
end_count = len(filenames)
for entry in filenames:
#want
dsift.process_image_dsift(entry, 'tmp.sift', 20, 10, resize=(200, 200))
l, d = sift.read_features_from_file('tmp.sift')
# if debug:
# cv2.destroyAllWindows()
#
# #used to draw on original image instead of resized image
# im = array(Image.open(entry))
# tmp_im = cv2.resize(im, (200,200))
# tmp_im = cv2.cvtColor(tmp_im, cv2.COLOR_BGR2GRAY)
# sift.plot_features(tmp_im, l, True)
#
# show()
image_vector.append(d.flatten().tolist())
os.remove('tmp.sift')
os.remove('tmp.frame')
if not PGM:
os.remove('tmp.pgm')
if out_file != None:
f = open(out_file, 'a')
if class_label == 1:
vector = '+1'
else:
vector = '-1'
vec_length = len(image_vector[ctr])
for i, dim in zip(xrange(vec_length), xrange(1, vec_length + 1)):
vector += ' ' + str(dim) + ':' + str(image_vector[ctr][i])
f.write(vector)
f.write('\n')
f.close()
ctr += 1
count += 1
#sys.stdout.write('\r%i/%i' % (count, end_count))
#sys.stdout.flush()
return image_vector
from PIL import Image
from numpy import *
from pylab import *
import os
from scipy import *
import sift,dsift
dsift.process_image_dsift('video/Fhist.jpg','video/FDsift.sift',30,15,True,resize=(200,200))
l,d = sift.read_features_from_file('video/FDsift.sift')
im = Image.open('video/Fhist.jpg')
im = im.resize((200,200))
im = array(im)
gray()
sift.plot_features(im,l,False)
show()
# -*- coding: utf-8 -*-
"""
Created on Tue Sep 13 21:37:23 2016
@author: user
"""
from PIL import Image
from pylab import *
from numpy import *
import dsift, sift
dsift.process_image_dsift('../pcv_data/data/empire.jpg', 'empire.sift', 90, 40,
True)
l, d = sift.read_features_from_file('empire.sift')
im = array(Image.open('../pcv_data/data/empire.jpg'))
sift.plot_features(im, l, True)
show()
class_2 = pickle.load(f)
labels = pickle.load(f)
# 最初の点についてテストしてみる
print model.classify(class_1[0])
# 描画用関数の定義
def classify(x,y,model=model):
return np.array([model.classify([xx,yy]) for (xx,yy) in zip(x,y)])
# 分類の境界を描画する
imtools.plot_2D_boundary([-6,6,-6,6],[class_1,class_2],classify,[1,-1])
# 8.1.2 画像特徴量としての密なＳＩＦＴ
dsift.process_image_dsift('empire.jpg','empire.sift',90,40,True)
l,d = sift.read_features_from_file('empire.sift')
im = np.array(Image.open('empire.jpg'))
sift.plot_features(im,l,True)
# 8.1.3 画像の分類－手話認識
def get_imlist_ppm(path):
return [os.path.join(path, f) for f in os.listdir(path) if f.endswith('.ppm')]
imlist = get_imlist_ppm('train')
# imlist = get_imlist_ppm('test')
# 画像を固定サイズ(50×50)で処理する
for filename in imlist:
featfile = filename[:-3] + 'dsift'
def dsift_features_generator(imlist):
for img in imlist:
featfile = img[:-3] + 'dsift'
dsift.process_image_dsift(img, featfile, 10, 5, resize=(50, 50))
test = []
for l in ls:
imgs = glob.glob(os.path.join(s, l, 'uniform', '*'))
train += imgs[::2]
test += imgs[1::2]
shutil.rmtree(d)
def md(p):
try: os.makedirs(p)
except: pass
md(os.path.join(d, 'train'))
md(os.path.join(d, 'test'))
for p in train: os.symlink(p, os.path.join(d, 'train', os.path.basename(p)))
for p in test: os.symlink(p, os.path.join(d, 'test', os.path.basename(p)))
# Compute dsift descriptors.
imlist = glob.glob(d + '/*/*.ppm')
dsiftlist = [os.path.splitext(im)[0] + '.dsift' for im in imlist]
# About 10s:
#for im, dim in zip(imlist, dsiftlist):
# dsift.process_image_dsift(im, dim, 10, 5, resize=(50, 50))
# About 7s:
from multiprocessing.dummy import Pool as ThreadPool
pool = ThreadPool(8)
pool.map(lambda x: dsift.process_image_dsift(x[0], x[1], 10, 5, resize=(50,50)),
zip(imlist, dsiftlist))
pool.close()
pool.join()
features = array(features)
# create labels
labels = [int(featfile.split('\\')[-1][0:3])/50 for featfile in featlist]
return features,array(labels)
imlist = []
path = 'C:\\Users\\RubyLyu\\PycharmProjects\\PictureDB\\data'
imlist=[os.path.join(path,f) for f in os.listdir(path) if f.endswith('.jpg')]
print imlist
for filename in imlist:
featfile = filename[:-3]+'dsift'
dsift.process_image_dsift(filename,featfile,10,5,resize=(150,200))
imlist = []
path = 'C:\\Users\\RubyLyu\\PycharmProjects\\PictureDB\\src\\data'
imlist=[os.path.join(path,f) for f in os.listdir(path) if f.endswith('.jpg')]
for filename in imlist:
featfile = filename[:-3]+'dsift'
dsift.process_image_dsift(filename,featfile,10,5,resize=(150,200))
features,labels = read_features_labels('C:\\Users\\RubyLyu\\PycharmProjects\\PictureDB\\data')
#print features
test_features,test_labels = read_features_labels('C:\\Users\\RubyLyu\\PycharmProjects\\PictureDB\\src\\data')
features = map(list,features)
def dsift_features_generator(imlist):
for img in imlist:
featfile = img[:-3]+'dsift'
dsift.process_image_dsift(img, featfile, 10, 5, resize=(50, 50))
from PIL import Image from pylab import * import dsift, sift f = '/Users/thakis/Downloads/data/empire.jpg' dsift.process_image_dsift(f, f + '.sift', 90, 40, True) loc, desc = sift.read_features_from_file(f + '.sift') im = array(Image.open(f)) sift.plot_features(im, loc, True) show()
if not load_model or not os.path.isfile(load_model):
print "No/invalid load model specified"
print_error()
# 3. Extract Features
imlist = [os.path.join(dirname, f) for f in os.listdir(dirname) if f.endswith('.jpg')] # create a list of images
for filename in imlist:
featfile = filename[:-3] + typeFeats
featfile = featfile.replace(" ", "")
try:
if not os.path.isfile(featfile):
dsift.process_image_dsift(filename, featfile, 10, 5, resize=(100, 100)) # process images at fixed size (100, 100)
except Exception:
pass
# 4. Run Classifier against
svm = pickle.load(open(load_model, "rb")) # load to model file
svm_test_features, filelist = read_feature_labels(dirname)
svm_test_result = svm.predict(svm_test_features)
result_dir = os.path.join('C:\Users', os.getenv('USERNAME'), 'Desktop\Result-Video')
if not os.path.isdir(result_dir):
os.mkdir(result_dir)
EXTENSIONS = [".jpg", ".bmp", ".png", ".pgm", ".tif", ".tiff"]
size = 20
step = 10
def get_categories(datasetpath):
cat_paths = [files for files in glob(datasetpath + "/*") if isdir(files)]
cat_paths.sort()
cats = [basename(cat_path) for cat_path in cat_paths]
return cats
def get_imgfiles(path):
all_files = []
all_files.extend([join(path, basename(fname)) for fname in glob(path + "/*") if splitext(fname)[-1].lower() in EXTENSIONS])
return all_files
dataset_path = '../flower_rec1/square_images128_dsift/test'
categories = get_categories(dataset_path)
for category in categories:
category_path = join(dataset_path, category)
image_file_list = get_imgfiles(category_path)
num_images = len(image_file_list)
feature_file_list = [image_file_list[i][:-3]+"dsift_{0}_{1}".format(size,step) for i in range(num_images)]
for i in range(num_images):
dsift.process_image_dsift(image_file_list[i], feature_file_list[i], size, step)
