def api_totals():
db, curs = database.connect()
total_attempts = database.get_total_attempts(db, curs)
total_ips = database.get_total_ips(db, curs)
db.close()
return "Total Attempts: {}<br>Total IPs: {}".format(
total_attempts, total_ips)
def submit_trait():
idx = request.form['traits']
password = request.form['pass']
if password == "scbd2019jaya":
db = connect()
cur = db.cursor()
cur.execute("SELECT * FROM ontologi_vote WHERE id = '{}'".format(idx))
res = []
for row in cur.fetchall():
res.append(row)
cur.execute("SELECT COUNT(*) FROM ontologi WHERE words = '{}'".format(
res[0][1]))
words_counts = cur.fetchone()[0]
cur.close()
if words_counts == 0:
cur = db.cursor()
cur.execute(
"INSERT INTO ontologi (words, traits, subtraits) VALUES ('{}', '{}', '{}')"
.format(res[0][1], res[0][2], res[0][2]))
db.commit()
cur.close()
cur = db.cursor()
cur.execute("DELETE FROM ontologi_vote WHERE texts = '{}'".format(
res[0][1]))
db.commit()
cur.close()
db.close()
flash('Sudah berhasil ditambahkan')
return redirect('list_new_traits')
else:
return redirect('view_submit_traits')
else:
redirect('view_submit_traits')
def get_lanes(self):
db = helper.connect()
cursor = db.cursor()
string = "SELECT OBJECT.ID FROM OBJECT,ANNOTATION WHERE ANNOTATION.ID=OBJECT.PID AND ANNOTATION.FILENAME='reference'"
cursor.execute(string)
res = cursor.fetchone()
lanes = []
cluster_ids = []
while res is not None:
list_x,list_y = [],[]
cursor_1 = db.cursor()
string = "SELECT OBJECT.ID,PT_CAMERA_COOR.T, PT_CAMERA_COOR.X,PT_CAMERA_COOR.Y,PT_CAMERA_COOR_ADD_OPTS.CLUSTER FROM PT_CAMERA_COOR,OBJECT,ANNOTATION,PT_CAMERA_COOR_ADD_OPTS WHERE OBJECT.ID = PT_CAMERA_COOR.PID AND ANNOTATION.ID=OBJECT.PID AND ANNOTATION.FILENAME='reference' AND OBJECT.ID=? AND PT_CAMERA_COOR_ADD_OPTS.ID = PT_CAMERA_COOR.ID ORDER BY CAST(PT_CAMERA_COOR.T AS UNSIGNED) ASC"
cursor_1.execute(string,[str(res[0])])
res_1 = cursor_1.fetchone()
cluster_id = None
while res_1 is not None:
list_x.append(float(res_1[2]))
list_y.append(float(res_1[3]))
if cluster_id == None:
cluster_id = res_1[4]
res_1 = cursor_1.fetchone()
smooth_list_x = s.smooth(np.asarray(list_x), window_len=3,window='flat')
smooth_list_y = s.smooth(np.asarray(list_y), window_len=3,window='flat')
theta_list = self.add_heading(smooth_list_x, smooth_list_y)
cluster_ids.append(cluster_id)
data_points_path = [(a,b) for a,b in zip(smooth_list_x,smooth_list_y)]
lanes.append(data_points_path)
res = cursor.fetchone()
db.close()
return lanes,cluster_ids
def add_trait():
words = request.form['words']
traits = request.form['traits']
if words != '' and traits != '':
words_clean = re.sub(r'[^a-z]', ' ', words.lower())
words_clean_fix = '_'.join(words_clean.split())
db = connect()
cur = db.cursor()
cur.execute(
"SELECT COUNT(*) FROM ontologi_vote WHERE texts = '{}'".format(
words_clean_fix))
words_counts = cur.fetchone()[0]
cur.close()
if words_counts == 0:
cur = db.cursor()
cur.execute(
"INSERT INTO ontologi_vote (texts, traits, upvote, downvote) VALUES ('{}', '{}', '0', '0')"
.format(words_clean_fix, traits))
db.commit()
cur.close()
db.close()
else:
print('Kata sudah ada')
return redirect('list_new_traits')
def api_recent(num):
db, curs = database.connect()
recents = database.get_recent(db, curs, num)
ret_str = ""
for row in recents:
ret_str += str(row)
ret_str += '<br>'
db.close()
return ret_str
def view_delete_traits():
db = connect()
cur = db.cursor()
res = []
cur.execute("SELECT * FROM ontologi_vote")
for row in cur.fetchall():
res.append(row)
cur.close()
db.close()
return render_template('view_delete_traits.html', traits=res)
def get_list_all_traits():
db = connect()
cur = db.cursor()
res = []
cur.execute("SELECT * FROM ontologi")
for row in cur.fetchall():
res.append(row)
cur.close()
db.close()
return render_template('list_all_traits.html', traits=res)
def downvote():
texts = request.form['texts']
db = connect()
cur = db.cursor()
cur.execute("UPDATE ontologi_vote \
SET downvote = downvote + 1 \
WHERE texts = '{}';".format(texts))
db.commit()
cur.close()
db.close()
return get_list_new_traits()
def result():
sent = request.form['input_user']
sent = sent.lower()
sent = re.sub(r'[^a-z]', ' ', sent)
tokens = sent.split()
list_traits = [
'Aggreeableness', 'Conscientiousness', 'Extraversion', 'Neuroticism',
'Openness'
]
list_freq = [0, 0, 0, 0, 0]
i = 0
db = connect()
res_table = []
while i < len(tokens):
temp = []
temp_traits = []
temp_subtraits = []
cur = db.cursor()
cur.execute("SELECT * FROM ontologi WHERE words like '%" + tokens[i] +
"%'")
for row in cur.fetchall():
temp.append(row[1])
temp_traits.append(row[2])
temp_subtraits.append(row[3])
maks = 0
trait = ''
subtrait = ''
words = ''
cur.close()
for k in range(len(temp)):
if re.sub('_', ' ', temp[k].lower()) in sent:
if len(temp[k].split('_')) > maks:
words = temp[k].split('_')
trait = temp_traits[k]
subtrait = temp_subtraits[k]
maks = len(temp[k].split('_'))
if maks > 0:
res_table.append([' '.join(words), trait, subtrait])
list_freq[list_traits.index(trait)] += 1
i += maks
else:
i += 1
db.close()
return render_template('result.html',
trait=list_traits,
freq=list_freq,
table_traits=res_table,
len=len(list_freq))
def sample(sampling_tuple, num_samples):
num_scenarios = list(range(1, sampling_tuple[0]))
num_scenarios.remove(sampling_tuple[1])
scene_2_sample_from = np.random.choice(num_scenarios)
db = helper.connect()
cursor = db.cursor()
string = "SELECT MIN(CAST(OBJECT.STARTFRAME AS UNSIGNED)),MAX(CAST(OBJECT.ENDFRAME AS UNSIGNED)) FROM OBJECT,ANNOTATION WHERE OBJECT.PID = ANNOTATION.ID AND ANNOTATION.ID =?"
cursor.execute(string, [str(scene_2_sample_from)])
res = cursor.fetchone()
min_frame, max_frame = int(res[0]), int(res[1])
db.close()
return np.random.randint(low=min_frame, high=max_frame,
size=num_samples), scene_2_sample_from
def delete_trait():
idx = request.form['traits']
password = request.form['pass']
if password == "scbd2019jaya":
db = connect()
cur = db.cursor()
cur.execute("DELETE FROM ontologi_vote WHERE id = '{}'".format(idx))
db.commit()
cur.close()
db.close()
return redirect('list_new_traits')
else:
redirect('view_delete_traits')
def main():
conn, cur = db.connect()
'''
UNCOMMENT to import training data
'''
db.import_training_data(conn, cur)
db.udpate_predict_training(conn, cur)
while (True):
line = raw_input('> ')
if line == 'exit':
db.close(conn, cur)
sys.exit()
print db.execute_cql(conn, cur, line)
def recent(num):
db, curs = database.connect()
recents = database.get_recent(db, curs, num)
db.close()
recent_map = folium.Map(location=[24.635246, 2.616971], zoom_start=2, tiles='CartoDB positron')
for row in recents:
coords = geo.get_coordinates(row[2])
if coords is None:
continue
folium.Marker(location=coords).add_to(recent_map)
recent_page = "<div style=\"height:50%; width:60%;\">" + str(recent_map.get_root().render()) + "</div>"
return recent_page #send_from_directory('res/html', 'recent.html')
def get_tj_ids(frames_2_sample, scene_2_sample_from):
db = helper.connect()
cursor = db.cursor()
tj_id_list, particle_id_list = [], []
for frames in frames_2_sample:
string = "SELECT PT_CAMERA_COOR.ID,OBJECT.ID FROM PT_CAMERA_COOR,OBJECT,ANNOTATION WHERE OBJECT.ID = PT_CAMERA_COOR.PID AND PT_CAMERA_COOR.T=? AND ANNOTATION.ID=OBJECT.PID AND ANNOTATION.ID=? ORDER BY CAST(PT_CAMERA_COOR.T AS UNSIGNED) ASC"
cursor.execute(string, [str(frames), str(scene_2_sample_from)])
res_1 = cursor.fetchone()
id_list_in_frame, particle_list_in_frame = [], []
while res_1 is not None:
id_list_in_frame.append(res_1[1])
particle_list_in_frame.append(res_1[0])
res_1 = cursor.fetchone()
tj_id_list.append(id_list_in_frame)
particle_id_list.append(particle_list_in_frame)
db.close()
return particle_id_list, tj_id_list
def apt_totalips():
db, curs = database.connect()
ret_str = str(database.get_total_ips(db, curs))
db.close()
return ret_str
def form_input():
db = helper.connect()
cursor = db.cursor()
string = "SELECT (SELECT COUNT(*) FROM OBJECT), (SELECT COUNT(*) FROM ANNOTATION);"
cursor.execute(string)
res = cursor.fetchone()
count, count_scenarios = res[0], res[1]
max_x, min_x, max_y, min_y = float('-inf'), float('inf'), float(
'-inf'), float('inf')
data_path, data_trajectory = dict(), dict()
particle_ids = []
ref_indexes = []
for i in range(1, count + 1):
cursor = db.cursor()
string = "SELECT PT_CAMERA_COOR.ID,PT_CAMERA_COOR.PID,PT_CAMERA_COOR.X,PT_CAMERA_COOR.Y,PT_CAMERA_COOR.T,OBJECT.NAME,ANNOTATION.ID,OBJECT.STARTFRAME,OBJECT.ENDFRAME FROM PT_CAMERA_COOR,OBJECT,ANNOTATION WHERE OBJECT.ID = PT_CAMERA_COOR.PID AND OBJECT.ID=? AND ANNOTATION.ID=OBJECT.PID AND ANNOTATION.ID NOT IN ('10') ORDER BY CAST(PT_CAMERA_COOR.T AS UNSIGNED) ASC"
cursor.execute(string, [i])
res_1 = cursor.fetchone()
if res_1 is None:
continue
h = int(res_1[6])
if h == 3:
ref_indexes.append(i)
last_x, last_y = None, None
list_x, list_y, list_type, list_x_v, list_y_v = [], [], [], [], []
traj_length = 0
while res_1 is not None:
t_id = res_1[0]
''' since the trajectories are ordered now, and will be merged again in order again, there is no need to keep track of the indices explicitly for particle ids'''
particle_ids.append(t_id)
traj_length = traj_length + 1
if max_x < float(res_1[2]):
max_x = float(res_1[2])
if min_x > float(res_1[2]):
min_x = float(res_1[2])
if max_y < float(res_1[3]):
max_y = float(res_1[3])
if min_y > float(res_1[3]):
min_y = float(res_1[3])
if last_x is not None and last_y is not None:
velocity_x, velocity_y = (float(res_1[2]) - last_x) * fps, (
float(res_1[3]) - last_y) * fps
list_x_v.append(velocity_x)
list_y_v.append(velocity_y)
last_x, last_y = float(res_1[2]), float(res_1[3])
list_x.append(float(res_1[2]))
list_y.append(float(res_1[3]))
type_str = str(res_1[5][:3])
if 'veh' == type_str:
list_type.append('1')
elif 'ped' == type_str:
list_type.append('0')
else:
print(False)
res_1 = cursor.fetchone()
list_x_v.insert(0, list_x_v[0])
list_y_v.insert(0, list_y_v[0])
list_x_a = list(
map(lambda v1, v2: (v2 - v1), list_x_v,
list_x_v[1:] + [list_x_v[-1]]))
list_y_a = list(
map(lambda v1, v2: (v2 - v1), list_y_v,
list_y_v[1:] + [list_y_v[-1]]))
'''smooth_list_x = s.smooth(np.asarray(list_x), window_len=10,window='flat')
smooth_list_y = s.smooth(np.asarray(list_y), window_len=10,window='flat')
smooth_list_x = s.smooth(smooth_list_x)
smooth_list_y = s.smooth(smooth_list_y)'''
''''z = [[e1,e2] for e1,e2 in zip(list_x,list_y)]
a = np.asarray(z)
smooth_list_x, smooth_list_y = s.gaussian_smoothing(a)'''
smooth_list_x, smooth_list_y = list_x, list_y
#print('size-x',len(list_x),len(smooth_list_x))
#print('size-y',len(list_y),len(smooth_list_y))
theta_list = add_heading(smooth_list_x, smooth_list_y)
theta_list = np.deg2rad(theta_list)
if not np.all(np.isfinite(np.asarray(theta_list))) or np.any(
np.isnan(np.asarray(theta_list))):
print(theta_list)
''' theta goes in as radians. omega goes in as sin(theta2) - sin(theta1)'''
omega_list = list(
map(lambda v1, v2: (v2 - v1), list(np.sin(theta_list)),
list(np.sin(theta_list))[1:] + [list(np.sin(theta_list))[-1]]))
'''print_stats(omega_list,i)'''
data_points_path = [
(a, b, c)
for a, b, c in zip(smooth_list_x, smooth_list_y, theta_list)
]
data_points_trajectory = [
(a, b, c, d, e, f, g1, h1, i1) for a, b, c, d, e, f, g1, h1, i1 in
zip(smooth_list_x, smooth_list_y, theta_list, list_x_v, list_y_v,
list_type, list_x_a, list_y_a, omega_list)
]
if h in data_path:
data_path[h].append(data_points_path)
data_trajectory[h].append(data_points_trajectory)
else:
data_path[h] = [data_points_path]
data_trajectory[h] = [data_points_trajectory]
db.close()
data_path_norm = data_path.copy()
''' Normalization'''
'''
for k in data_path_norm:
for i in range(len(data_path_norm[k])):
for j in range(len(data_path_norm[k][i])):
data_path_norm[k][i][j] = ((data_path_norm[k][i][j][0] - min_x ) / (max_x - min_x) , (data_path_norm[k][i][j][1] - min_y ) / (max_y - min_y) , (data_path_norm[k][i][j][2]) / 360)
'''
return data_path, data_trajectory, particle_ids
def starred(ip):
db, curs = database.connect()
marker_color = database.get_color(db, curs, ip)
db.close()
return '<html><body style="background:{};"></body></html>'.format(
marker_color)
def api_totalattempts():
db, curs = database.connect()
ret_str = str(database.get_total_attempts(db, curs))
db.close()
return ret_str
def get_a_data_batch():
density_max = 0
db = helper.connect()
cursor = db.cursor()
string = """SELECT PT_CAMERA_COOR.ID,PT_CAMERA_COOR.PID,PT_CAMERA_COOR.X,PT_CAMERA_COOR.Y,PT_CAMERA_COOR.T,PT_CAMERA_COOR_ADD_OPTS.THETA,PT_CAMERA_COOR_ADD_OPTS.X_V,PT_CAMERA_COOR_ADD_OPTS.Y_V,PT_CAMERA_COOR_ADD_OPTS.CLASS,ANNOTATION.ID,OBJECT.ID,PT_CAMERA_COOR_ADD_OPTS.CLUSTER
FROM PT_CAMERA_COOR,PT_CAMERA_COOR_ADD_OPTS,OBJECT,ANNOTATION WHERE OBJECT.ID = PT_CAMERA_COOR.PID AND ANNOTATION.ID=OBJECT.PID AND ANNOTATION.ID NOT IN ('8') AND PT_CAMERA_COOR_ADD_OPTS.ID = PT_CAMERA_COOR.ID
ORDER BY CAST(PT_CAMERA_COOR.T AS UNSIGNED) ASC"""
cursor.execute(string)
results = cursor.fetchall()
x = map(list, list(results))
x = list(x)
n = np.asarray(x, dtype=np.float32)
db.close()
env_batch = np.zeros(shape=(BATCH_SIZE, 5, 6, 7), dtype=np.float32)
env_batch_slice = np.zeros(shape=(5, 6, 7), dtype=np.float32)
target_batch = np.zeros(shape=(BATCH_SIZE, 2, 1), dtype=np.float32)
ego_batch = np.zeros(shape=(BATCH_SIZE, 7, 1), dtype=np.float32)
'''here :env_batch = np.zeros(shape=(BATCH_SIZE,1,7), dtype = np.float32)
here :env_batch_slice = np.zeros(shape=(1,7), dtype = np.float32)
here :target_batch = np.zeros(shape=(BATCH_SIZE,2,1) , dtype = np.float32)
here :ego_batch = np.zeros(shape=(BATCH_SIZE,7,1) , dtype = np.float32)'''
''' we need consecutive frames of the same annotation id having the same object ids
once we have these 2 frames, we create a image grid and a target grid
1. sample one scenario randomly between min and max but not in evaluation scenario
2. get t_id range in scenario from 1. sample a t_id
3. run 2 queries with (scenario id in 1 and t_id in 2) and (scenario id in 1 and t_id+1 in 2)
4. if list of object_ids in the result of both queries in 2 are same, use the result as consecutive frames. else from 2.
'''
fill_index = 0
max_density_index = 0
while fill_index < 20000:
found_it = False
while (not found_it):
#grid_data = np.zeros(shape=(GRID_SIZE,GRID_SIZE,4))
#grid_target = np.zeros(shape=(GRID_SIZE,GRID_SIZE,3))
''' Step 1'''
min_s, max_s = np.amin(n[:, 9]), np.amax(n[:, 9])
scenes = list(range(int(min_s), int(max_s) + 1))
scenes.remove(8)
scenario_id = np.random.choice(scenes)
''' Step 2'''
c1 = n[:, 9] == scenario_id
r = n[c1]
min, max = np.amin(r[:, 4]), np.amax(r[:, 4])
t_id = np.random.randint(min, max + 1)
''' Step 3'''
c2 = n[:, 4] == t_id
c3 = n[:, 4] == t_id + 1
c4 = np.logical_and(c1, c2)
c5 = np.logical_and(c1, c3)
r1 = n[c4]
r2 = n[c5]
''' Step 4'''
object_ids_r1 = r1[:, 10]
object_ids_r2 = r2[:, 10]
'''if len(object_ids_r1) != density_2_look_4:
found_it = False
continue'''
if np.array_equal(object_ids_r1, object_ids_r2):
for o_ids in object_ids_r1:
env_batch_slice = np.zeros(shape=(5, 6, 6),
dtype=np.float32)
'''here :env_batch_slice = np.zeros(shape=(1,6), dtype = np.float32)'''
#print('density is',len(object_ids_r1),' ego object id is',o_ids)
density_value = len(object_ids_r1)
if density_value > density_max:
density_max = density_value
max_density_index = fill_index
density_array = np.zeros(shape=(5, 6, 1), dtype=np.float32)
'''here :density_array = np.zeros(shape=(1,1), dtype = np.float32)'''
density_array.fill(density_value)
obj_ind = {0: 0, 1: 0, 2: 0, 3: 0, 4: 0}
for i in range(r1.shape[0]):
if r1[i, 10] == o_ids:
#print('ego obj id is',o_ids)
x, y = float(r1[i, 2]), float(r1[i, 3])
fea_vector = np.reshape(np.asarray([
x, y,
float(r1[i, 6]),
float(r1[i, 7]),
int(r1[i, 8]),
int(r1[i, 11]), density_value
]),
newshape=(7, 1))
ego_batch[fill_index] = np.copy(fea_vector)
for i2 in range(r2.shape[0]):
if r2[i2, 10] == o_ids:
#print('target entered for obj id',o_ids)
y_a = float(r2[i2, 7]) - float(r1[i, 7])
x_a = float(r2[i2, 6]) - float(r1[i, 6])
x_n, y_n = float(r2[i, 2]), float(r2[i, 3])
fea_vector = np.reshape(np.asarray(
[x_a, y_a]),
newshape=(2, 1))
target_batch[fill_index] = np.copy(
fea_vector)
else:
#print('added env entry for object id',r1[i,10])
x, y = float(r1[i, 2]), float(r1[i, 3])
cluster_id = int(r1[i, 11])
fea_vector = np.asarray([
x, y,
float(r1[i, 6]),
float(r1[i, 7]),
int(r1[i, 8]) + 10,
int(r1[i, 11])
])
if cluster_id < 5:
env_batch_slice[cluster_id,
obj_ind[cluster_id]] = np.copy(
fea_vector)
'''here : env_batch_slice[0] = np.copy(fea_vector)'''
obj_ind[cluster_id] = obj_ind[cluster_id] + 1
#env_batch_slice = np.sort(env_batch_slice,axis=1)
env_batch[fill_index] = np.concatenate(
(env_batch_slice, density_array), axis=2)
'''here : env_batch[fill_index] = np.concatenate((env_batch_slice,density_array),axis=1)'''
fill_index = fill_index + 1
print(fill_index)
print('------')
found_it = True
# else:
# print(False)
#print('max density is ',density_max)
return env_batch[0:fill_index, ], ego_batch[0:fill_index, ], target_batch[
0:fill_index, ], (max_density_index, density_max)
# Draw maps using Folium and write to an html file.
# Author: N. Beckstead
#
# TODO: Debug when a marker field is None.
# TODO: Draw heatmap using log_mapper.attempts table.
# TODO: Optimize by looking up geo data first. Then call make_marker()
#
import folium
from folium import plugins
import db_helper as database
import geo_helper as geo
import server_vars
MARKER_RADIUS = 4
db, curs = database.connect()
#
# Main map function. Draw all maps.
#
def draw():
markers_map = folium.Map(location=[24.635246, 2.616971],
zoom_start=3,
tiles='CartoDB dark_matter')
heatmap = folium.Map(location=[24.635246, 2.616971],
zoom_start=3,
tiles='CartoDB positron')
markers_map = make_markersmap(markers_map)
def get_a_evaluation_data_batch(for_dbn):
db = helper.connect()
cursor = db.cursor()
string = """SELECT PT_CAMERA_COOR.ID,PT_CAMERA_COOR.PID,PT_CAMERA_COOR.X,PT_CAMERA_COOR.Y,PT_CAMERA_COOR.T,PT_CAMERA_COOR_ADD_OPTS.THETA,PT_CAMERA_COOR_ADD_OPTS.X_V,PT_CAMERA_COOR_ADD_OPTS.Y_V,PT_CAMERA_COOR_ADD_OPTS.CLASS,ANNOTATION.ID,OBJECT.ID,PT_CAMERA_COOR_ADD_OPTS.CLUSTER
FROM PT_CAMERA_COOR,PT_CAMERA_COOR_ADD_OPTS,OBJECT,ANNOTATION WHERE OBJECT.ID = PT_CAMERA_COOR.PID AND ANNOTATION.ID=OBJECT.PID AND ANNOTATION.ID IN ('8') AND PT_CAMERA_COOR_ADD_OPTS.ID = PT_CAMERA_COOR.ID
ORDER BY CAST(PT_CAMERA_COOR.T AS UNSIGNED) ASC"""
cursor.execute(string)
results = cursor.fetchall()
x = map(list, list(results))
x = list(x)
n = np.asarray(x, dtype=np.float32)
db.close()
env_batch = np.zeros(shape=(BATCH_SIZE, 5, 6, 7), dtype=np.float32)
target_batch = np.zeros(shape=(BATCH_SIZE, 2, 1), dtype=np.float32)
ego_batch = np.zeros(shape=(BATCH_SIZE, 7, 1), dtype=np.float32)
t_start, t_end = int(n[0, 4]), int(n[n.shape[0] - 1, 4] + 1)
fill_index = 0
for t_id in range(t_start, t_end - 2):
c2 = n[:, 4] == t_id
c3 = n[:, 4] == t_id + 1
r1 = n[c2]
r2 = n[c3]
object_ids_r1 = r1[:, 10]
object_ids_r2 = r2[:, 10]
'''print('looking for density :',density_2_look_4)
if len(object_ids_r1) != density_2_look_4:
found_it = False
continue'''
for o_ids in object_ids_r1:
density_value = len(object_ids_r1)
obj_ind = 0
env_batch_slice = np.zeros(shape=(5, 6, 6), dtype=np.float32)
'''here : env_batch_slice = np.zeros(shape=(1,6), dtype = np.float32)'''
density_array = np.zeros(shape=(5, 6, 1), dtype=np.float32)
'''here : density_array = np.zeros(shape=(1,1), dtype = np.float32)'''
density_array.fill(len(object_ids_r1))
for i in range(r1.shape[0]):
if r1[i, 10] == o_ids:
x, y = float(r1[i, 2]), float(r1[i, 3])
if for_dbn is False:
fea_vector = np.reshape(np.asarray([
x, y,
float(r1[i, 6]),
float(r1[i, 7]),
int(r1[i, 8]),
int(r1[i, 11]), density_value
]),
newshape=(7, 1))
else:
fea_vector = np.reshape(np.asarray([
x, y,
float(r1[i, 5]),
float(r1[i, 6]),
float(r1[i, 7]),
int(r1[i, 8]),
int(r1[i, 11])
]),
newshape=(7, 1))
ego_batch[fill_index] = np.copy(fea_vector)
for i2 in range(r2.shape[0]):
if r2[i2, 10] == o_ids:
print('target entered for obj id', o_ids)
y_a = float(r2[i2, 7]) - float(r1[i, 7])
x_a = float(r2[i2, 6]) - float(r1[i, 6])
fea_vector = np.reshape(np.asarray([x_a, y_a]),
newshape=(2, 1))
target_batch[fill_index] = np.copy(fea_vector)
else:
#print('added env entry for object id',r1[i,10])
x, y = float(r1[i, 2]), float(r1[i, 3])
cluster_id = int(r1[i, 11])
fea_vector = np.asarray([
x, y,
float(r1[i, 6]),
float(r1[i, 7]),
int(r1[i, 8]) + 10,
int(r1[i, 11])
])
if cluster_id < 5:
env_batch_slice[cluster_id,
obj_ind] = np.copy(fea_vector)
''' here : env_batch_slice[0] = np.copy(fea_vector)'''
obj_ind = obj_ind + 1
if for_dbn:
env_batch_slice = np.sort(env_batch_slice, axis=1)
env_batch[fill_index] = np.concatenate(
(env_batch_slice, density_array), axis=2)
'''here : env_batch[fill_index] = np.concatenate((env_batch_slice,density_array),axis=1)'''
fill_index = fill_index + 1
print(fill_index)
print('------')
print('size of eval set :', fill_index)
return env_batch[0:fill_index, ], ego_batch[0:fill_index, ], target_batch[
0:fill_index, ], (0, 0)
