def load_from_file(load_filename, memory_size=10000):
"""Loads and returns a GroupStore object from file."""
store = GroupStore(load_filename, memory_size)
file = open(load_filename, "r")
i = 0
for line in file:
input = line.split(",", 1)
modularity = float(input[0])
grouping = Grouping.create_from_string(input[1], None, modularity)
store.groups[i] = grouping
i = i + 1
if not (i < memory_size):
i = 0
store.update_stats()
# Clear the buffer...
for j in xrange(i, memory_size):
store.groups[j] = None
store.update_stats()
# Now clear the rest...
for j in xrange(0, i):
store.groups[i] = None
return store
def load_from_file(load_filename, memory_size = 10000):
"""Loads and returns a GroupStore object from file."""
store = GroupStore(load_filename, memory_size)
file = open(load_filename, "r")
i = 0
for line in file:
input = line.split(",",1)
modularity = float(input[0])
grouping = Grouping.create_from_string(input[1],None,modularity)
store.groups[i] = grouping
i = i+1
if not (i < memory_size):
i = 0
store.update_stats()
# Clear the buffer...
for j in xrange(i,memory_size):
store.groups[j] = None
store.update_stats()
# Now clear the rest...
for j in xrange(0, i):
store.groups[i] = None
return store
def _get_msg_group(self):
msg_group = []
dataset_list = ['eth', 'eth', 'ucy', 'ucy', 'ucy']
dataset_idx_list = [0, 1, 0, 1, 2]
#dataset_list = ['eth']
#dataset_idx_list = [0]
for i in range(len(dataset_list)):
dataset = dataset_list[i]
dataset_idx = dataset_idx_list[i]
msg = Message()
data = DataLoader(dataset, dataset_idx, self.fps)
msg = data.update_message(msg)
gp = Grouping(msg, self.history)
msg = gp.update_message(msg)
msg_group.append(msg)
return msg_group
def startHenshu(basepath):
func = {
'G1': getInfoImage,
'G2': getInfoZip,
'G3': getInfoZip,
'G4': getInfoZip,
'G5': getInfoZip,
'G6': getInfoZip
}
tmpDic = dict()
for (path, dir, filenames) in os.walk(basepath):
tmpList = list()
for filename in filenames:
# 확장자 축출 @[1:] .이후 문자를 가져옴
fext = os.path.splitext(filename)[-1][1:].lower()
vDic = dict()
fullpath = os.path.join(path, filename)
dirname = path.split('\\')[-1]
# 분류를 위한 평가요소는 확장자, 파일명(키워드포함), 폴더명(키워드포함)
# 결과 ext는 확장자평가, cat는 파일명/폴더명내 포함된 키워드에 의한 평가
# 파일명에 확장자 제외
fn = filename.split('.')[0]
ext, cat = grp.getKey(fext, fn, dirname)
# 어느 것에도 기타로 분류될 경우, 대상외로 함.
if ext == 'G6' and cat == 'C5':
continue
try:
vDic = func[ext](fullpath)
except:
# 에러가 발생하더라도 계속 수행
log.printlog("ERR:{}".format(fullpath))
continue
else:
# 정상적인 경우 집계절차
vDic['filename'] = filename
vDic['ext'], vDic['cat'] = ext, cat
tmpList.append(vDic)
if tmpList:
tmpDic[dirname] = tmpList
log.printlog(str(dirname))
# print(tmpDic)
with io.open('./fileCollection/result.txt', 'w', encoding='utf-8') as f:
json.dump(tmpDic, f, ensure_ascii=False, indent=1)
log.printlog("********Henshu complete*******")
def run_test():
######################################################
#
# Runs a couple sanity checks on the network:
#
# 0 0
# |\ /|
# | 0--0 |
# |/ \|
# 0 0
#
# with the obvious correct grouping followed
# by a slightly non-optimal grouping.
#
# Results should be Q1 ~ .357, Q2 ~ .122
#
######################################################
network = create_test_network()
print "A = ", network
group1 = Grouping.create_from_vector([1, 1, 1, 0, 0, 0], network)
print "S1 = ", group1
group2 = Grouping.create_from_vector([1, 1, 1, 1, 0, 0], network)
print "S2 = ", group2
group3 = Grouping.create_from_vector([1, 1, 0, 1, 0, 0], network)
print "S3 = ", group3
Q1 = group1.get_modularity()
Q2 = group2.get_modularity()
Q3 = group3.get_modularity()
print "Modularities:"
print "Q_1 = ", Q1
print "Q_2 = ", Q2
print "Q_3 = ", Q3
if ((Q1 - 0.357) < 0.001) and ((Q2 - 0.122) < 0.001):
print "Test passed!"
else:
print "*** Test Failed! ***"
return False
print "Trying point patch method..."
pointpatch = PointPatch(3, 0, 1)
Q4 = pointpatch.get_modularity(group1, network)
print "Q_4 = ", Q4
print group1.get_nodes()
print group1.get_groups()
pointpatch.modify(group1)
print group1.get_nodes()
print group1.get_groups()
Q5 = group1.calculate_modularity(network)
print "Q_5 = ", Q5
if Q5 - Q2 < 0.00001:
print "Test passed!"
else:
print "*** Test Failed! ***"
return False
return True
def create_graphml_file(self, filename, grouping=None):
"""Creates a graphml file that can be used to generate a visualization
of the network using the optional partition to color the nodes."""
file = open(filename + ".graphml", "w")
if grouping == None:
grouping = Grouping.create_from_vector([0] * self.number_of_nodes)
n = self.number_of_nodes
m = self.number_of_edges
nodes = grouping.get_nodes()
non_empty_groups = grouping.get_non_empty_groups()
c = len(non_empty_groups)
A = self.adjacency_matrix
print n, m, c, len(nodes), len(colors)
# set palettes...
if c <= len(colors):
if c == 1:
palette = ["#FFFFFF"]
else:
palette = graphml_colors
else:
palette = map(lambda x: "gray" + str(x), range(0, 100))
# Preamble stuff...
file.write(
'<?xml version="1.0" encoding="UTF-8" standalone="no"?>\n<graphml xmlns="http://graphml.graphdrawing.org/xmlns/graphml" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:y="http://www.yworks.com/xml/graphml" xsi:schemaLocation="http://graphml.graphdrawing.org/xmlns/graphml http://www.yworks.com/xml/schema/graphml/1.0/ygraphml.xsd">\n<key for="node" id="d0" yfiles.type="nodegraphics"/>\n<key attr.name="description" attr.type="string" for="node" id="d1"/>\n<key for="edge" id="d2" yfiles.type="edgegraphics"/>\n<key attr.name="description" attr.type="string" for="edge" id="d3"/>\n<key for="graphml" id="d4" yfiles.type="resources"/>\n'
)
# Graph header
file.write(
"""<graph edgedefault="directed" id="G" parse.edges="%d" parse.nodes="%d" parse.order="free">\n""" % (m, n)
)
# Now we write the nodes...
for i in xrange(0, n):
file.write(
"""<node id="n%d">
<data key="d0">
<y:ShapeNode>
<y:Geometry height="30.0" width="30.0" x="80.56138057743618" y="0.0"/>
<y:Fill color="%s" transparent="false"/>
<y:BorderStyle color="#000000" type="line" width="1.0"/>
<y:NodeLabel alignment="center" autoSizePolicy="content" fontFamily="Dialog" fontSize="12" fontStyle="plain" hasBackgroundColor="false" hasLineColor="false" height="18.701171875" modelName="internal" modelPosition="c" textColor="#000000" visible="true" width="33.0" x="-1.5" y="5.6494140625">%d</y:NodeLabel>
<y:Shape type="ellipse"/>
</y:ShapeNode>
</data>
<data key="d1"/>
</node>\n"""
% (i, palette[non_empty_groups.index(nodes[i])], i)
)
current_edge = 0
# Now we do the edges...
for i in range(0, n):
for j in range(0, i + 1):
if A[i, j] == 1:
file.write(
"""<edge id="e%d" source="n%d" target="n%d">
<data key="d2">
<y:PolyLineEdge>
<y:Path sx="0.0" sy="0.0" tx="0.0" ty="0.0"/>
<y:LineStyle color="#000000" type="line" width="1.0"/>
<y:Arrows source="none" target="none"/>
<y:BendStyle smoothed="false"/>
</y:PolyLineEdge>
</data>
<data key="d3"/>
</edge>\n"""
% (current_edge, i, j)
)
current_edge += 1
# close off tags...
file.write(
""" </graph>
<data key="d4">
<y:Resources/>
</data>
</graphml>"""
)
file.close()
print "Reading data file..."
df = pd.read_csv(file_path, sep=",", encoding='utf-8')
df['attr_val_pairs'] = df['attr_val_pairs'].apply(lambda x: json.loads(x))
df['variant_criterias'] = df['variant_criterias'].apply(
lambda x: json.loads(x))
print "Reading product data..."
product_data = list(df.itertuples(index=False))
print "Getting items..."
items = utils.get_unique_items_pt(product_data, item_type)
print "Creating groups..."
grp_instance = Grouping(items)
grp_instance.init_groups()
print "Cluster entropy score = ", grp_instance.get_clustering_scores()
groups1 = grp_instance.auto_groups if group_type == 'auto' else grp_instance.true_groups
print "Reading excluded attribute list..."
with open('excluded_attr_list.txt', 'rb') as x_attr:
excluded_attrs = x_attr.readlines()
excluded_attrs = [x.strip() for x in excluded_attrs]
excluded_attrs = set(excluded_attrs)
print "Getting variants..."
get_variants(items, groups1, excluded_attrs)
dataset_idx = 0 history = 16 frame_idx = 500 group_idx = 20 # Initialize a message msg = Message() # Initialize dataloader data = DataLoader(dataset, dataset_idx) # Update the message msg = data.update_message(msg) # Initialize grouping gp = Grouping(msg, 16) # Update the message msg = gp.update_message(msg) # This shows what group ids are in this frame print(msg.video_labels_matrix[frame_idx]) # Initialize group shape generation gs_gen = GroupShapeGeneration(msg) vertices, pedidx = gs_gen.generate_group_shape(frame_idx, group_idx) # The returned vertices for the group shape print(vertices) print(pedidx) # We can also draw it on an image (blank canvas in this case) canvas = np.zeros((msg.frame_height, msg.frame_width, 3), dtype=np.uint8)
def run_test():
######################################################
#
# Runs a couple sanity checks on the network:
#
# 0 0
# |\ /|
# | 0--0 |
# |/ \|
# 0 0
#
# with the obvious correct grouping followed
# by a slightly non-optimal grouping.
#
# Results should be Q1 ~ .357, Q2 ~ .122
#
######################################################
network = create_test_network()
print "A = ", network
group1 = Grouping.create_from_vector([1, 1, 1, 0, 0, 0], network)
print "S1 = ", group1
group2 = Grouping.create_from_vector([1, 1, 1, 1, 0, 0], network)
print "S2 = ", group2
group3 = Grouping.create_from_vector([1, 1, 0, 1, 0, 0], network)
print "S3 = ", group3
Q1 = group1.get_modularity()
Q2 = group2.get_modularity()
Q3 = group3.get_modularity()
print "Modularities:"
print "Q_1 = ", Q1
print "Q_2 = ", Q2
print "Q_3 = ", Q3
if ((Q1 - .357) < .001) and ((Q2 - .122) < .001):
print "Test passed!"
else:
print "*** Test Failed! ***"
return False
print "Trying point patch method..."
pointpatch = PointPatch(3, 0, 1)
Q4 = pointpatch.get_modularity(group1, network)
print "Q_4 = ", Q4
print group1.get_nodes()
print group1.get_groups()
pointpatch.modify(group1)
print group1.get_nodes()
print group1.get_groups()
Q5 = group1.calculate_modularity(network)
print "Q_5 = ", Q5
if Q5 - Q2 < .00001:
print "Test passed!"
else:
print "*** Test Failed! ***"
return False
return True
def create_graphml_file(self, filename, grouping=None):
"""Creates a graphml file that can be used to generate a visualization
of the network using the optional partition to color the nodes."""
file = open(filename + ".graphml", "w")
if grouping == None:
grouping = Grouping.create_from_vector([0] * self.number_of_nodes)
n = self.number_of_nodes
m = self.number_of_edges
nodes = grouping.get_nodes()
non_empty_groups = grouping.get_non_empty_groups()
c = len(non_empty_groups)
A = self.adjacency_matrix
print n, m, c, len(nodes), len(colors)
# set palettes...
if c <= len(colors):
if c == 1:
palette = ["#FFFFFF"]
else:
palette = graphml_colors
else:
palette = map(lambda x: "gray" + str(x), range(0, 100))
# Preamble stuff...
file.write(
'<?xml version="1.0" encoding="UTF-8" standalone="no"?>\n<graphml xmlns="http://graphml.graphdrawing.org/xmlns/graphml" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:y="http://www.yworks.com/xml/graphml" xsi:schemaLocation="http://graphml.graphdrawing.org/xmlns/graphml http://www.yworks.com/xml/schema/graphml/1.0/ygraphml.xsd">\n<key for="node" id="d0" yfiles.type="nodegraphics"/>\n<key attr.name="description" attr.type="string" for="node" id="d1"/>\n<key for="edge" id="d2" yfiles.type="edgegraphics"/>\n<key attr.name="description" attr.type="string" for="edge" id="d3"/>\n<key for="graphml" id="d4" yfiles.type="resources"/>\n'
)
# Graph header
file.write(
"""<graph edgedefault="directed" id="G" parse.edges="%d" parse.nodes="%d" parse.order="free">\n"""
% (m, n))
# Now we write the nodes...
for i in xrange(0, n):
file.write("""<node id="n%d">
<data key="d0">
<y:ShapeNode>
<y:Geometry height="30.0" width="30.0" x="80.56138057743618" y="0.0"/>
<y:Fill color="%s" transparent="false"/>
<y:BorderStyle color="#000000" type="line" width="1.0"/>
<y:NodeLabel alignment="center" autoSizePolicy="content" fontFamily="Dialog" fontSize="12" fontStyle="plain" hasBackgroundColor="false" hasLineColor="false" height="18.701171875" modelName="internal" modelPosition="c" textColor="#000000" visible="true" width="33.0" x="-1.5" y="5.6494140625">%d</y:NodeLabel>
<y:Shape type="ellipse"/>
</y:ShapeNode>
</data>
<data key="d1"/>
</node>\n""" % (i, palette[non_empty_groups.index(nodes[i])], i))
current_edge = 0
# Now we do the edges...
for i in range(0, n):
for j in range(0, i + 1):
if A[i, j] == 1:
file.write("""<edge id="e%d" source="n%d" target="n%d">
<data key="d2">
<y:PolyLineEdge>
<y:Path sx="0.0" sy="0.0" tx="0.0" ty="0.0"/>
<y:LineStyle color="#000000" type="line" width="1.0"/>
<y:Arrows source="none" target="none"/>
<y:BendStyle smoothed="false"/>
</y:PolyLineEdge>
</data>
<data key="d3"/>
</edge>\n""" % (current_edge, i, j))
current_edge += 1
# close off tags...
file.write(""" </graph>
<data key="d4">
<y:Resources/>
</data>
</graphml>""")
file.close()
def predict(self, positions, velocities):
params = self._load_parameters()
position_array = []
velocity_array = []
num_people = len(positions)
if num_people == 0:
raise Exception('People Needed!')
seq_length = len(positions[0])
pred_seq_length = 8
#gp = Grouping(self.msg, seq_length)
#self.msg = gp.update_message(self.msg)
for i in range(num_people):
position_array.append(positions[i][-1])
velocity_array.append(velocities[i][-1])
labels = Grouping.grouping(position_array, velocity_array, params)
all_labels = np.unique(labels)
num_groups = len(all_labels)
all_pred_img_sequences = []
#gsg = GroupShapeGeneration(self.msg)
for curr_label in all_labels:
group_positions = []
group_velocities = []
for i, l in enumerate(labels):
if l == curr_label:
group_positions.append(positions[i])
group_velocities.append(velocities[i])
vertice_sequence = []
for i in range(seq_length):
frame_positions = []
frame_velocities = []
for j in range(len(group_positions)):
frame_positions.append(group_positions[j][i])
frame_velocities.append(group_velocities[j][i])
vertices = GroupShapeGeneration.draw_social_shapes(frame_positions,
frame_velocities)
vertice_sequence.append(vertices)
dgs = DrawGroupShape(self.msg)
dgs.set_center(vertice_sequence)
dgs.set_aug(angle=0)
img_sequence = []
for i, v in enumerate(vertice_sequence):
canvas = np.zeros((self.msg.frame_height, self.msg.frame_width, 3), dtype=np.uint8)
img = dgs.draw_group_shape(v, canvas, center=True, aug=False)
img_sequence.append(img)
pimg = ProcessImage(self.msg, img_sequence)
for i, img in enumerate(img_sequence):
img_sequence[i] = pimg.process_image(img, debug=False)
pred_img_sequence = self._predict_sequence(img_sequence, pred_seq_length)
group_pred_img_sequence = []
for i, img in enumerate(pred_img_sequence):
#img = np.round(np.repeat(img, 3, axis=2)) * 255
img = np.round(np.repeat(img, 3, axis=2))
pred_img = pimg.reverse_process_image(img, debug=True)
pred_img = dgs.reverse_move_center_img(pred_img)
group_pred_img_sequence.append(pred_img[:, :, 0])
all_pred_img_sequences.append(group_pred_img_sequence)
fnl_pred_img_sequence = []
for i in range(pred_seq_length):
canvas = np.zeros((self.msg.frame_height, self.msg.frame_width), dtype=np.uint8)
for j in range(num_groups):
img = all_pred_img_sequences[j][i]
#img = np.round(img)
canvas += img
fnl_pred_img_sequence.append(np.clip(canvas, 0, 1))
return fnl_pred_img_sequence
