def start_kmeans(args):
imlist = imtools.get_imlist(args.data_dir)
imnbr = len(imlist) # get the number of images
print("The number of images is %d" % imnbr)
# Create matrix to store all flattened images
immatrix = np.array([np.array(cv2.imread(imname)).flatten() for imname in imlist], 'f')
# PCA reduce dimension
V, S, immean = pca.pca(immatrix)
immatrix = np.array([np.array(cv2.imread(im)).flatten() for im in imlist])
immatrix_src = np.array([np.array(cv2.imread(im)) for im in imlist])
# project on the 40 first PCs
immean = immean.flatten()
projected = np.array([dot(V[:40], immatrix[i] - immean) for i in range(imnbr)])
# k-means
projected = whiten(projected)
centroids, distortion = kmeans(projected, args.num_class)
code, distance = vq(projected, centroids)
output_path = os.path.join(args.output_dir, 'kmeans_result')
if not os.path.exists(output_path):
os.mkdir(output_path)
# plot clusters
for k in range(args.num_class):
ind = where(code == k)[0]
print("class:", k, len(ind))
os.mkdir(os.path.join(output_path, str(k)))
i = rdm.randint(0,len(ind))
cv2.imwrite(os.path.join(os.path.join(output_path, str(k)), str(i) + '.jpg'), immatrix_src[ind[i]])
def save_vis(filename_list, number_list, pred_pca, output_hcluster, output_PCA):
imlist = filename_list
imnbr = len(imlist)
# Load images, run PCA.
immatrix = array(pred_pca)
V, S, immean = pca.pca(immatrix)
# Project on 2 PCs.
projected = array([dot(V[[0, 1]], immatrix[i] - immean) for i in range(imnbr)])
# height and width
h, w = 1200, 1200
# create a new image with a white background
img = Image.new('RGB', (w, h), (255, 255, 255))
draw = ImageDraw.Draw(img)
# draw axis
draw.line((0, h/2, w, h/2), fill=(255, 0, 0))
draw.line((w/2, 0, w/2, h), fill=(255, 0, 0))
# scale coordinates to fit
scale = abs(projected).max(0)
scaled = floor(array([(p/scale) * (w/2 - 20, h/2 - 20) + (w/2, h/2)
for p in projected])).astype(int)
# paste thumbnail of each image
for i in range(imnbr):
nodeim = Image.open(imlist[i])
nodeim.thumbnail((25, 25))
ns = nodeim.size
box = (scaled[i][0] - ns[0] // 2, scaled[i][1] - ns[1] // 2,
scaled[i][0] + ns[0] // 2 + 1, scaled[i][1] + ns[1] // 2 + 1)
img.paste(nodeim, box)
tree = hcluster.hcluster(projected)
hcluster.draw_dendrogram(tree,imlist,filename=output_hcluster)
for i, num in enumerate(number_list):
if num < 8:
color1 = mod(370*num,255)
color2 = mod(170*num,255)
color3 = mod(270*num,255)
draw.text((scaled[i][0],scaled[i][1]),str(num),fill = (color1,color2,color3))
else:
draw.text((scaled[i][0],scaled[i][1]),str(num),fill = (255,0,0))
figure()
imshow(img)
axis('off')
img.save(output_PCA)
show()
def inference(mnist):
x = tf.placeholder(dtype=tf.float32, shape=[None, 784])
x_image = tf.reshape(x, [-1, 28, 28, 1])
global_step = tf.Variable(0, trainable=False)
W_conv1 = weight_variable([3, 3, 1, 128])
b_conv1 = bias_variable([128])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
W_conv2 = weight_variable([3, 3, 128, 100])
b_conv2 = bias_variable([100])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2)
# W_conv3 = weight_variable([3, 3, 128, 64])
# b_conv3 = bias_variable([64])
# h_conv3 = tf.nn.relu(conv2d(h_pool2, W_conv3) + b_conv3)
# h_pool3 = max_pool_2x2(h_conv3)
# W_conv4 = weight_variable([5, 5, 64, 32])
# b_conv4 = bias_variable([32])
# h_conv4 = tf.nn.relu(conv2d(h_pool3, W_conv4) + b_conv4)
# h_pool4 = max_pool_2x2(h_conv4)
h_pool4_flat = tf.reshape(h_pool2, [-1, 7*7*100])
# gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.4)
# config = tf.ConfigProto(gpu_options=gpu_options)
# config.gpu_options.allow_growth = True
with tf.Session() as sess:
tf.global_variables_initializer().run()
xs = mnist.test.images
ys = mnist.test.labels
extract_features = sess.run(h_pool4_flat, feed_dict={x: xs})
print(extract_features.shape)
V, S, immean = pca.pca(extract_features)
immean = immean.flatten()
imnbr = 10000
projected = array([dot(V[:32], extract_features[i] - immean) for i in range(imnbr)])
centroids, distortion = kmeans(projected, 20)
code, distance = vq(projected, centroids)
tools.calculate_acc(code, ys, 10, 20)
return extract_features
def main(args):
print(args)
# 根据指定总帧数和间隔抽取帧
frames, frame_ids = get_frames(args.input, args.frame_num, args.frame_interval, args.max_size)
print(len(frames), len(frame_ids), frames[0].shape)
immatrix = np.array([x.flatten() for x in frames]).astype(np.float32)
print(immatrix.shape)
# PCA降维
V, S, immean = pca.pca(immatrix)
print(V.shape, S.shape, immean.shape)
imnbr = len(frames)
projected = np.array([np.dot(V[:40],immatrix[i]-immean) for i in range(imnbr)])
# k-means
projected = whiten(projected)
print(projected.shape, type(projected))
centroids,distortion = kmeans(projected, args.key_frame_num)
code,distance = vq(projected,centroids)
print(code.shape, distance.shape)
print(code)
print(distance)
# 使用Kmeans进行聚类
# kmeans = KMeans(n_clusters=args.key_frame_num, random_state=0).fit(np.array(frames))
centers = []
clusters = [[] for i in range(args.key_frame_num)]
ids = [[] for i in range(args.key_frame_num)]
for i in range(len(frames)):
clusters[code[i]].append(distance[i])
ids[code[i]].append(i)
print(ids)
print(clusters)
for i in range(args.key_frame_num):
index = np.argsort(clusters[i])
j = index[0]
print(index)
index = ids[i][j]
print(index)
centers.append((frames[index], index))
if not os.path.isdir(args.out_dir):
os.makedirs(args.out_dir)
basename = os.path.basename(args.input)
for i, (frame, index) in enumerate(centers):
cv2.imwrite(os.path.join(args.out_dir, basename + '_%d_%d.jpg' % (index,i)), frame)
# -*- coding: utf-8 -*-
from PCV.tools import imtools, pca
from PIL import Image, ImageDraw
from pylab import *
imlist = imtools.get_imlist('../data/selectedfontimages/a_selected_thumbs')
imnbr = len(imlist)
# Load images, run PCA.
immatrix = array([array(Image.open(im)).flatten() for im in imlist], 'f')
V, S, immean = pca.pca(immatrix)
# Project on 2 PCs.
projected = array([dot(V[[0, 1]], immatrix[i] - immean) for i in range(imnbr)]) # P131 Fig6-3左图
#projected = array([dot(V[[1, 2]], immatrix[i] - immean) for i in range(imnbr)]) # P131 Fig6-3右图
# height and width
h, w = 1200, 1200
# create a new image with a white background
img = Image.new('RGB', (w, h), (255, 255, 255))
draw = ImageDraw.Draw(img)
# draw axis
draw.line((0, h/2, w, h/2), fill=(255, 0, 0))
draw.line((w/2, 0, w/2, h), fill=(255, 0, 0))
# scale coordinates to fit
scale = abs(projected).max(0)
scaled = floor(array([(p/scale) * (w/2 - 20, h/2 - 20) + (w/2, h/2)
for p in projected])).astype(int)
# -*- coding: utf-8 -*-
from PCV.tools import imtools, pca
from PIL import Image, ImageDraw
from pylab import *
from scipy.cluster.vq import *
imlist = imtools.get_imlist('tmp/test/')
imnbr = len(imlist)
# Load images, run PCA.
immatrix = array([array(Image.open(im)).flatten() for im in imlist], 'f')
V, S, immean = pca.pca(immatrix)
projected = array([dot(V[[0, 1]], immatrix[i] - immean) for i in range(imnbr)])
n = len(projected)
# compute distance matrix
S = array([[sqrt(sum((projected[i] - projected[j])**2)) for i in range(n)]
for j in range(n)], 'f')
# create Laplacian matrix
rowsum = sum(S, axis=0)
D = diag(1 / sqrt(rowsum))
I = identity(n)
L = I - dot(D, dot(S, D))
# compute eigenvectors of L
U, sigma, V = linalg.svd(L)
k = 20
# create feature vector from k first eigenvectors
# by stacking eigenvectors as columns
features = array(V[:k]).T
# k-means
def PCA_Kmeans(train_path,result_path):
# Uses sparse pca codepath.
# imlist = imtools.get_imlist('C:/Users/Zqc/Desktop/PCV-book-data/data/selectedfontimages/a_selected_thumbs/')
# falter_image("D:/datebase/test")
imlist = imtools.get_imlist(train_path)
# print(imlist)
# 获取图像列表和他们的尺寸
im = array(Image.open(imlist[0])) # open one image to get the size
m, n = im.shape[:2] # get the size of the images
imnbr = len(imlist) # get the number of images
print("The number of images is %d" % imnbr)
# Create matrix to store all flattened images
immatrix = array([array(Image.open(imname)).flatten() for imname in imlist], 'f')
# PCA降维
V, S, immean = pca.pca(immatrix)
# 保存均值和主成分
# f = open('./a_pca_modes.pkl', 'wb')
# f = open('C:/Users/Zqc/Desktop/PCV-book-data/data/selectedfontimages/a_pca_modes.pkl', 'wb')
f = open(train_path + "\\a_pca_modes.pkl", 'wb')
pickle.dump(immean, f)
pickle.dump(V, f)
f.close()
# get list of images
# imlist = imtools.get_imlist('C:/Users/Zqc/Desktop/PCV-book-data/data/selectedfontimages/a_selected_thumbs/')
print("go")
imlist = imtools.get_imlist(train_path)
imnbr = len(imlist)
# load model file
# with open('C:/Users/Zqc/Desktop/PCV-book-data/data/selectedfontimages/a_pca_modes.pkl','rb') as f:
with open(train_path + "a_pca_modes.pkl", 'rb') as f:
immean = pickle.load(f)
V = pickle.load(f)
# create matrix to store all flattened images
immatrix = array([array(Image.open(im)).flatten() for im in imlist], 'f')
# project on the 40 first PCs
immean = immean.flatten()
projected = array([dot(V[:30], immatrix[i] - immean) for i in range(imnbr)])
print("PCA")
# k-means
projected = whiten(projected)
centroids, distortion = kmeans(projected, 370)
code, distance = vq(projected, centroids)
filepath = result_path
# plot clusters
for k in range(370):
tempath = filepath + str(k)
os.makedirs(tempath)
ind = where(code == k)[0]
for i in range(len(ind)):
io.imsave(tempath + "\\" + str(i) + ".png", immatrix[ind[i]].reshape((100, 100)).astype(np.uint8),
cmap="gray")
print 'training data is:', features.shape, len(labels)
# read test data
####################
test_features,test_labels = read_gesture_features_labels('../data/hand_gesture/test/')
print 'test data is:', test_features.shape, len(test_labels)
classnames = unique(labels)
nbr_classes = len(classnames)
if False:
# reduce dimensions with PCA
from PCV.tools import pca
V,S,m = pca.pca(features)
# keep most important dimensions
V = V[:50]
features = array([dot(V,f-m) for f in features])
test_features = array([dot(V,f-m) for f in test_features])
if True:
# test kNN
k = 1
knn_classifier = knn.KnnClassifier(labels,features)
res = array([knn_classifier.classify(test_features[i],k) for i in range(len(test_labels))]) # TODO kan goras battre
if False:
print 'training data is:', features.shape, len(labels)
# read test data
####################
test_features, test_labels = read_gesture_features_labels(
'../data/hand_gesture/test/')
print 'test data is:', test_features.shape, len(test_labels)
classnames = unique(labels)
nbr_classes = len(classnames)
if False:
# reduce dimensions with PCA
from PCV.tools import pca
V, S, m = pca.pca(features)
# keep most important dimensions
V = V[:50]
features = array([dot(V, f - m) for f in features])
test_features = array([dot(V, f - m) for f in test_features])
if True:
# test kNN
k = 1
knn_classifier = knn.KnnClassifier(labels, features)
res = array([
knn_classifier.classify(test_features[i], k)
for i in range(len(test_labels))
]) # TODO kan goras battre
else:
temp = "%d: None" % (j)
str += temp + "| "
summ += max_count
if step % 1000 == 0:
print("in cluster %d:" % i)
print(str + "|| MAX PERCENT: %d: %g of %d" % (f, ma, len(part)))
print("TOTAL ACC %g" % (summ/len(label)))
mnist = input_data.read_data_sets("mnist_data", one_hot=False)
xs = mnist.test.images
ys = mnist.test.labels
V, S, immean = pca.pca(xs)
f = open('./mnist_test_pca_modes.pkl', 'wb')
pickle.dump(immean, f)
pickle.dump(V, f)
f.close()
immean = immean.flatten()
imnbr = 10000
projected = array([dot(V[:32], xs[i]-immean) for i in range(imnbr)])
# k-means
projected = whiten(projected)
centroids, distortion = kmeans(projected, 15)
