def configure(self, rf):
"""Configure the module according to the configuration file
Args:
rf (string): name of the configuration file of the module.
This file must be include in the folder pointing to the yarp
contexts. By default: rf = 'default.ini'
"""
self.handlerPort.open("/ErgonomicAssessmentModule")
self.attach(self.handlerPort)
self.in_port_posture = yarp.BufferedPortBottle()
self.in_port_posture.open('/ergo_pred/posture:i')
path_config_posture = rf.find('posture_config').toString()
self._posture = HumanPosture(path_config_posture)
path_config_ergo = rf.find('ergo_config').toString()
self._ergo_assessment = ErgoAssessment(path_config_ergo)
self.list_out_port_ergo = dict()
for ergo_score in self._ergo_assessment.get_list_score():
self.list_out_port_ergo[ergo_score] = yarp.BufferedPortBottle()
self.list_out_port_ergo[ergo_score].open("/ergo_pred/" +
ergo_score.lower() + ':o')
return True
def reduce_skeleton(data, num_frame, list_score):
score_total = []
human_posture = HumanPosture('config/mapping_joints.json')
ergo_assessment = ErgoAssessment('config/reba_config.json')
human_posture.update_posture(data)
for score in list_score:
score_total.append(ergo_assessment[score])
return score_total
def prepare_data(self, metric, input_data):
if metric == 'posture':
human_pose = HumanPosture('config/mapping_joints.json')
pool = mp.Pool(mp.cpu_count())
result_objects = [
pool.apply_async(human_pose.update_posture, args=(data, i))
for i, data in enumerate(input_data)
]
data_eval = np.asarray([r.get() for r in result_objects])
pool.close()
pool.join()
elif metric == 'position':
end_effectors = [
'Right Hand', 'Left Hand', 'Right Foot', 'Left Foot'
]
pool = mp.Pool(mp.cpu_count())
result_objects = [
pool.apply_async(skeleton.update_posture, args=(data, True, i))
for i, data in enumerate(input_data)
]
data_eval = [r.get() for r in result_objects]
pool.close()
pool.join()
elif metric == 'ergo_score':
self.ergo_assessment = ErgoAssessment('config/rula_config.json')
self.list_ergo_score = self.ergo_assessment.get_list_score_name()
self.list_ergo_score.sort()
if (type(input_data) == torch.Tensor):
for i, data in enumerate(input_data[0:1]):
data_eval = tools.compute_sequence_ergo(
data, i, self.list_ergo_score)
elif (type(input_data) == np.ndarray):
pool = mp.Pool(mp.cpu_count())
result_objects = [
pool.apply_async(tools.compute_sequence_ergo,
args=(data, i, self.list_ergo_score))
for i, data in enumerate(input_data)
]
data_eval = [r.get() for r in result_objects]
data_eval = np.asarray(data_eval)
pool.close()
pool.join()
else:
data_eval = np.copy(input_data)
return data_eval
def compute_sequence_ergo(ergo_assessment, data, num_frame, list_score,
config_path):
score_total = []
human_posture = HumanPosture(config_path + 'config/mapping_joints.json')
if len(data) == human_posture.get_dim_reduce():
human_posture.set_posture(data)
elif len(data) == human_posture.get_dim_input():
human_posture.update_posture(data)
ergo_assessment.compute_ergo_scores(human_posture)
for score in list_score:
score_total.append(ergo_assessment[score])
return score_total
def __init__(self):
self.app = pg.mkQApp()
pg.setConfigOption('background', 'w')
pg.setConfigOption('foreground', 'k')
self.view = pg.PlotWidget()
self.view.resize(800, 600)
self.view.setWindowTitle('Ergonomic score in latent space')
self.view.setAspectLocked(True)
self.view.show()
self.port = yarp.BufferedPortBottle()
self.port.open('/plot_latentspace')
metric = 'jointAngle'
ergo_name = ['TABLE_REBA_C']
size_latent = 2
dx = 0.1
loss = [[]]
autoencoder = []
all_score = []
all_size = []
type_data = []
path_src = "/home/amalaise/Documents/These/code/ergo_prediction/ergonomic_assessment/src/"
path = path_src + "save/AE/" + metric + "/" + str(size_latent) + '/'
list_files = [f for f in os.listdir(path) if os.path.isfile(os.path.join(path, f))]
list_files.sort()
file = list_files[0]
with open(path + file, 'rb') as input:
autoencoder = pickle.load(input)
input_data = autoencoder.get_data_test()
data_output, encoded_data, score = autoencoder.test_model(input_data)
score = autoencoder.evaluate_model(input_data, data_output, metric)
Max = np.max(encoded_data, axis = 0)
Min = np.min(encoded_data, axis = 0)
Mean = np.mean(encoded_data, axis = 0)
# Compute ergo score
ergo_assessment = ErgoAssessment(path_src + 'config/rula_config.json')
list_ergo_score = ergo_assessment.get_list_score_name()
list_ergo_score.sort()
reduce_posture = HumanPosture(path_src + 'config/mapping_joints.json')
posture = Skeleton('dhm66_ISB_Xsens.urdf')
self.X = np.arange(0.0, 1.0+dx, dx)
self.ergo_grid = np.zeros((len(self.X), len(self.X)))
for i, data_x in enumerate(self.X):
for j, data_y in enumerate(self.X):
x = np.zeros((1,size_latent))
x[0, 0] = data_x
x[0, 1] = data_y
decoded_data = autoencoder.decode_data(x)
if metric == 'posture':
whole_body = reduce_posture.reduce2complete(decoded_data[0])
posture.update_posture(whole_body)
else:
posture.update_posture(decoded_data[0])
ergo_score = tools.compute_sequence_ergo(decoded_data[0], 0, ergo_name, path_src)[0]
if ergo_score == 1:
ergo_score = 1
elif 1 < ergo_score < 5:
ergo_score = 2
elif 4 < ergo_score < 6:
ergo_score = 3
else:
ergo_score = 4
self.ergo_grid[j,i] = ergo_score
self.flag = 0
self.plot_latent_space()
# list_files.sort()
# file = list_files[0]
file = '/home/amalaise/Documents/These/code/ergo_prediction/ergonomic_assessment/src/save/AE/jointAngle/2/autoencoder_2_0.pkl'
with open(file, 'rb') as input:
autoencoder = pickle.load(input)
# data_output, encoded_data, score = autoencoder.test_model(input_data)
# Compute ergo score
ergo_assessment = ErgoAssessment('config/' + config_ergo + '_config.json')
list_ergo_score = ergo_assessment.get_list_score_name()
list_ergo_score.sort()
reduce_posture = HumanPosture('config/mapping_joints.json')
posture = Skeleton('dhm66_ISB_Xsens.urdf')
X = np.arange(0.0, 1.0 + dx, dx)
path_src = "/home/amalaise/Documents/These/code/ergo_prediction/ergonomic_assessment/src/"
pd_ergo_grid = pd.read_csv(config_ergo + '_grid.csv', index_col=False)
ergo_grid = np.asarray(pd_ergo_grid)
fig = plt.figure()
# ax1 = Axes3D(fig)
# ax1 = fig.add_subplot(122, projection='3d')
# plt.imshow(ergo_grid, extent = [X[0] , X[-1]-X[0], X[-1]-X[0] , X[0]])
# plt.imshow(ergo_grid)
def __init__(self, parent=None, config_ergo='rula', **kargs):
pg.GraphicsWindow.__init__(self, **kargs)
self.setParent(parent)
timer = pg.QtCore.QTimer(self)
timer.timeout.connect(self.update)
timer.start(10) # number of seconds (every 1000) for next update
self.view = pg.PlotWidget(self)
self.view.resize(350, 350)
self.view.setWindowTitle('Ergonomic score in latent space')
self.view.setAspectLocked(True)
self.view.show()
self.port = yarp.BufferedPortBottle()
self.port.open('/plot_latentspace/' + config_ergo)
yarp.Network.connect('/AE/latent:o', self.port.getName())
metric = 'jointAngle'
path_src = "/home/amalaise/Documents/These/code/ergo_prediction/ergonomic_assessment/src/"
self.config_ergo = config_ergo
if self.config_ergo == 'rula':
ergo_name = ['RULA_SCORE']
ergo_assessment = ErgoAssessment(path_src +
'config/rula_config.json')
else:
ergo_name = ['TABLE_REBA_C']
ergo_assessment = ErgoAssessment(path_src +
'config/reba_config.json')
size_latent = 2
dx = 0.02
loss = [[]]
autoencoder = []
all_score = []
all_size = []
type_data = []
path = path_src + "save/AE/" + metric + "/" + str(size_latent) + '/'
list_files = [
f for f in os.listdir(path)
if os.path.isfile(os.path.join(path, f))
]
list_files.sort()
file = list_files[0]
with open(path + file, 'rb') as input:
autoencoder = pickle.load(input)
input_data = autoencoder.get_data_test()
data_output, encoded_data, score = autoencoder.test_model(input_data)
score = autoencoder.evaluate_model(input_data, data_output, metric)
Max = np.max(encoded_data, axis=0)
Min = np.min(encoded_data, axis=0)
Mean = np.mean(encoded_data, axis=0)
# Compute ergo score
list_ergo_score = ergo_assessment.get_list_score_name()
list_ergo_score.sort()
reduce_posture = HumanPosture(path_src + 'config/mapping_joints.json')
posture = Skeleton('dhm66_ISB_Xsens.urdf')
self.X = np.arange(0.0, 1.0 + dx, dx)
pd_ergo_grid = pd.read_csv(self.config_ergo + '_grid.csv',
index_col=False)
self.ergo_grid = np.asarray(pd_ergo_grid)
ergo_color_grid = np.zeros(
(len(self.ergo_grid), len(self.ergo_grid), 3))
print(np.shape(ergo_color_grid))
for i in range(len(self.ergo_grid)):
for j in range(len(self.ergo_grid)):
if config_ergo == 'reba':
if self.ergo_grid[i, j] == 1:
ergo_color_grid[i, j] = [0, 1, 0]
elif self.ergo_grid[i, j] == 2:
ergo_color_grid[i, j] = [0.3, 0.7, 0]
elif self.ergo_grid[i, j] == 3:
ergo_color_grid[i, j] = [0.5, 0.5, 0]
elif self.ergo_grid[i, j] == 4:
ergo_color_grid[i, j] = [0.7, 0.3, 0]
else:
if self.ergo_grid[i, j] == 2:
ergo_color_grid[i, j] = [0, 1, 0]
elif self.ergo_grid[i, j] == 3:
ergo_color_grid[i, j] = [0.3, 0.7, 0]
elif self.ergo_grid[i, j] == 4:
ergo_color_grid[i, j] = [0.5, 0.5, 0]
elif self.ergo_grid[i, j] == 5:
ergo_color_grid[i, j] = [0.7, 0.3, 0]
self.ergo_grid = ergo_color_grid
# self.ergo_grid = np.zeros((len(self.X), len(self.X)))
# for i, data_x in enumerate(self.X):
# for j, data_y in enumerate(self.X):
# x = np.zeros((1,size_latent))
# x[0, 0] = data_x
# x[0, 1] = data_y
# decoded_data = autoencoder.decode_data(x)
# if metric == 'posture':
# whole_body = reduce_posture.reduce2complete(decoded_data[0])
# posture.update_posture(whole_body)
# else:
# posture.update_posture(decoded_data[0])
# ergo_score = tools.compute_sequence_ergo(ergo_assessment, decoded_data[0], 0, ergo_name, path_src)[0]
# if self.config_ergo == 'reba':
# if ergo_score == 1:
# ergo_score = 1
# elif 1 < ergo_score < 5:
# ergo_score = 2
# elif 4 < ergo_score < 6:
# ergo_score = 3
# else:
# ergo_score = 4
# self.ergo_grid[j,i] = ergo_score
self.flag = 0
self.plot_latent_space()
type_data = []
path = "save/AE/" + metric + "/" + str(size_latent) + '/'
list_files = [
f for f in os.listdir(path) if os.path.isfile(os.path.join(path, f))
]
list_files.sort()
file = list_files[0]
print(file)
with open(path + file, 'rb') as input:
autoencoder = pickle.load(input)
input_data = autoencoder.get_data_test()
data_output, encoded_data, score = autoencoder.test_model(input_data)
reduce_posture = HumanPosture('config/mapping_joints.json')
data_input_whole = np.zeros((len(input_data), 66))
data_output_whole = np.zeros((len(input_data), 66))
if metric == 'posture':
for i, data in enumerate(data_output):
data_input_whole[i] = reduce_posture.reduce2complete(input_data[i])
data_output_whole[i] = reduce_posture.reduce2complete(data)
else:
data_input_whole = input_data
data_output_whole = data_output
posture = Skeleton('dhm66_ISB_Xsens.urdf')
class ErgonomicAssessmentModule(yarp.RFModule):
"""This module compute online the ergonomic scores of postures
related to worksheets from industry
Example:
>>> python online_assessment.py --from ergo_config.ini
Attributes:
in_port_posture (BufferedPortBottle): Input port. It is used to read
the posture data.
list_out_port_ergo (Dict{BufferedPortBottle}): List of output ports.
They send the ergonomic value of each measure included in the
configuration file.
"""
def __init__(self):
yarp.RFModule.__init__(self)
self.handlerPort = yarp.Port()
def configure(self, rf):
"""Configure the module according to the configuration file
Args:
rf (string): name of the configuration file of the module.
This file must be include in the folder pointing to the yarp
contexts. By default: rf = 'default.ini'
"""
self.handlerPort.open("/ErgonomicAssessmentModule")
self.attach(self.handlerPort)
self.in_port_posture = yarp.BufferedPortBottle()
self.in_port_posture.open('/ergo_pred/posture:i')
path_config_posture = rf.find('posture_config').toString()
self._posture = HumanPosture(path_config_posture)
path_config_ergo = rf.find('ergo_config').toString()
self._ergo_assessment = ErgoAssessment(path_config_ergo)
self.list_out_port_ergo = dict()
for ergo_score in self._ergo_assessment.get_list_score():
self.list_out_port_ergo[ergo_score] = yarp.BufferedPortBottle()
self.list_out_port_ergo[ergo_score].open("/ergo_pred/" +
ergo_score.lower() + ':o')
return True
def close(self):
"""Close all the ports used by the module
"""
self.in_port_posture.close()
for ergo_score in self.list_out_port_ergo:
self.list_out_port_ergo[ergo_score].close()
self.handlerPort.close()
return True
def updateModule(self):
"""Read the data of posture and send the updated ergonomic scores
"""
b_in = self.in_port_posture.read(True)
data = b_in.toString().split(' ')
value = list(map(float, data))
self._posture.update_posture(value)
self._ergo_assessment.compute_ergo_scores(self._posture)
for ergo_score in self.list_out_port_ergo:
b_out = self.list_out_port_ergo[ergo_score].prepare()
b_out.clear()
b_out.addDouble(self._ergo_assessment[ergo_score])
self.list_out_port_ergo[ergo_score].write()
return True