def next_batch_test2():
train = DataSet('UCF', 'train1', 25)
# profile.run("next_batch_test2()")
# profile.run("next_batch_test()")
# for i in range(480):
# res = train.next_batch(500)
# print(len(res),str(train.index_in_epoch_video), train.index_frame)
#
# for i in range(10):
# print(random.randint(0, 1))
def next_batch_test():
train = DataSet('UCF', 'train1', 25)
images, labels = train.next_batch(500)
# test.append(tmp2)
# tmp = numpy.array(test)
# print(tmp.shape)
arr = []
for i in range(2):
arr.append(numpy.arange(10))
#
# for i in range(10):
# numpy.random.shuffle(arr[0])
# numpy.random.shuffle(arr[1])
# print(arr)
train = DataSet('UCF', 'train1', 25)
begin_time = time.time()
images, labels = train.next_batch(2)
print(time.time() - begin_time, 's')
print(images.shape)
print(labels.shape)
print(labels[0])
plt.imshow(images[0])
plt.show()
def next_batch_test():
train = DataSet('UCF', 'train1', 25)
images, labels = train.next_batch(500)
def next_batch_test2():
train = DataSet('UCF', 'train1', 25)
import time
import tensorflow as tf
from scripts import DataSet
import numpy
import urllib.request
# import os
# os.environ["CUDA_VISIBLE_DEVICES"] = '0'
sys.path.append("D:/graduation_project/workspace/models/research/slim")
from scripts import my_vgg_16 as vgg
slim = tf.contrib.slim
checkpoint_path = 'D:/graduation_project/checkpoints/vgg_16.ckpt'
txtName = 'D:/graduation_project/workspace/dataset/UCF101_train_test_splits/train1_feature_files/vgg16_fc7/'
train = DataSet.DataSet('UCF', 'train3', 25, label_type='notonehot')
total_frame = train.total_frame
myinput = tf.placeholder(tf.float32, [None, 224, 224, 3])
logits, _ = vgg.vgg_16(myinput, num_classes=1000, is_training=False, jud='fc7')
init = slim.assign_from_checkpoint_fn(
checkpoint_path, slim.get_variables_to_restore(include=["vgg_16"]))
beginTime = time.time()
pre = 0
post = 9999
status = True
with tf.Session() as sess:
init(sess)
for k in range(total_frame):
if status is True:
nowTxtName = txtName + str(pre) + '_' + str(post) + '.txt'
probabilities = tf.nn.softmax(logits)
print(slim.get_model_variables())
# 从checkpoint读入网络权值
init_fn = slim.assign_from_checkpoint_fn(checkpoints_dir, slim.get_model_variables('vgg_16'))
show_detail = False
right_count = 0
with tf.Session() as sess:
# 加载权值
begin_time = time.time()
init_fn(sess)
print('Loading: %.2fs' % (time.time()-begin_time))
test = DataSet.Test('UCF101', 'test1', 10)
begin_time = time.time()
for k in range(len(file_list)):
file = file_list[k]
ground_true = (file.split('.')[0]).split('_')[1]
input_batch = test.test_an_video(source_folder + file)
result = probabilities.eval(feed_dict={input: input_batch})
if show_detail:
print('===================== ' + str(k) + ' ==================')
seq = []
for x in range(result.shape[0]):
prob = result[x, 0:]
sorted_inds = [i[0] for i in sorted(enumerate(-prob),
key=lambda x: x[1])]
if show_detail:
def makeMatrix(self, ds, filterMin, filterMax, reorder):
# getting data from FileLoader and creating the right format
dataSetSort = DataSetSort()
if reorder == 0:
nodes = ds.getDoubleList(filterMin, filterMax, True)
elif reorder == 1:
dsCopy = DataSet(ds.getNodes())
dsCopy.makeUndirectionalAdd()
nodes = dsCopy.getDoubleList(filterMin, filterMax, True)
elif reorder == 2:
dataSetSort.DesConnectionSort(ds)
nodes = ds.getDoubleList(filterMin, filterMax, True)
elif reorder == 3:
dataSetSort.DesStrengthSort(ds)
nodes = ds.getDoubleList(filterMin, filterMax, True)
elif reorder == 4:
nodes = ds.distanceMatrix(True)
elif reorder == 5:
dataSetSort.robinsonSort(ds)
nodes = ds.getDoubleList(filterMin, filterMax, True)
names = ds.getNames()
yNames = names.copy()
yNames.reverse()
df = pd.DataFrame(
nodes,
columns=yNames,
index=names)
df.index.name = 'X'
df.columns.name = 'Y'
# Prepare data.frame in the right format
df = df.stack().rename("value").reset_index()
# Making the plot html file
output_file("matrixPlot.html")
# Creating the array containing the colors
colorList = []
i = 0
while i < 256:
color = wc.rgb_to_hex((255-(i-10), 255-(i-20), 255-(i-30)))
colorList.append(color)
i = i + 10
#colorList = [(23, 165, 137), (19, 141, 117), (40, 180, 99), (36, 113, 163)
# , (31, 97, 141), (17, 122, 101), (46, 134, 193), (34, 153, 84)]#,(202, 111, 30), (186, 74, 0)]
#colors = []
#for i in colorList:
# color = wc.rgb_to_hex(i)
# colors.append(color)
colors = colorList
# This part maps the colors at intervals
mapper = LinearColorMapper(
palette=colors, low=df.value.min(), high=df.value.max())
# Creating the figure
p = figure(
plot_width=500,
plot_height=500,
x_range=list(df.X.drop_duplicates()),
y_range=list(df.Y.drop_duplicates()),
# Adding a toolbar
#toolbar_location="right",
#tools="hover,pan,box_zoom,undo,redo,reset,save",
x_axis_location="above")
node_hover_tool = HoverTool(tooltips=[("Name X Axis", "@X"), ("Name Y Axis", "@Y"), ("Relation Strength", "@value")])
# plot.add_tools(node_hover_tool, BoxZoomTool(), ResetTool())
p.add_tools(node_hover_tool, TapTool(), BoxSelectTool(), BoxZoomTool(), UndoTool(), RedoTool())
p.toolbar_location = 'right'
# Create rectangle for heatmap
p.rect(
x="X",
y="Y",
width=1,
height=1,
source=ColumnDataSource(df),
line_color=None,
fill_color=transform('value', mapper))
# Add legend
color_bar = ColorBar(
color_mapper=mapper,
location=(0, 0),
ticker=BasicTicker(desired_num_ticks=len(colors)))
if len(list(df.X.drop_duplicates())) > p.plot_width / 10:
p.xaxis.visible = False
else:
pass
if len(list(df.Y.drop_duplicates())) > p.plot_height / 10:
p.yaxis.visible = False
else:
pass
p.xaxis.major_label_orientation = 'vertical'
p.add_layout(color_bar, 'right')
return p
import sys
import time
import tensorflow as tf
from scripts import DataSet
import numpy
import urllib.request
# import os
# os.environ["CUDA_VISIBLE_DEVICES"] = '0'
sys.path.append("D:/graduation_project/workspace/models/research/slim")
from scripts import my_vgg_16 as vgg
slim = tf.contrib.slim
checkpoint_path = 'D:/graduation_project/checkpoints/vgg_16.ckpt'
new_model_checkpoint_path = 'D:/graduation_project/workspace/checkpoints'
train = DataSet.DataSet('UCF', 'train1', 25)
input, label = train.next_batch(2)
logits, _ = vgg.vgg_16(input, num_classes=1000, is_training=False, jud='fc7')
init1 = slim.assign_from_checkpoint_fn(
checkpoint_path, slim.get_variables_to_restore(include=["vgg_16"]))
# init1 = slim.assign_from_checkpoint_fn(new_model_checkpoint_path + '\\model.ckpt-107', slim.get_variables("vgg_16"))
# init2 = slim.assign_from_checkpoint_fn(new_model_checkpoint_path + '\\model.ckpt-108', slim.get_variables("vgg_16"))
with tf.Session() as sess:
# init1(sess)
# fc8_biases = slim.get_variables("vgg_16/fc8/biases")
# fc7_biases = slim.get_variables("vgg_16/fc7/biases")
# print(fc8_biases)
# print('fc8 biases pre: ', sess.run(fc8_biases[0:10]))
# print('fc7 biases pre: ', sess.run(fc7_biases[0:10]))