def __init__(self, ork_params, action_name):
# create the plasm that will run the detection
self.plasm = create_plasm(ork_params)
self.plasm.configure_all()
rospy.loginfo('ORK server configured')
# the results or the object recognition pipeline
self.recognition_result = None
topics = ['recognized_object_array']
for sink in ork_params.values():
if 'recognized_object_array_topic' in sink:
topics.append(sink['recognized_object_array_topic'])
# subscribe to the output of the detection pipeline
for topic in topics:
rospy.Subscriber(topic, RecognizedObjectArray, self.callback_recognized_object_array)
rospy.loginfo('Subscribed to the ' + topic + ' topic.')
# look for a cell that contains a cropper to select the ROI
self.cropper = None
for cell in self.plasm.cells():
if 'crop_enabled' in cell.params:
self.cropper = cell
# actionlib stuff
if action_name and action_name != '""':
self.server = actionlib.SimpleActionServer(action_name, ObjectRecognitionAction, self.execute, False)
else:
self.server = actionlib.SimpleActionServer('recognize_objects', ObjectRecognitionAction, self.execute, False)
self.server.start()
rospy.loginfo('ORK server started')
def __init__(self, ork_params):
# create the plasm that will run the detection
self.plasm = create_plasm(ork_params)
self.plasm.configure_all()
rospy.loginfo('ORK server configured')
# the results or the object recognition pipeline
self.recognition_result = None
topics = ['recognized_object_array']
for sink in ork_params.values():
if 'recognized_object_array_topic' in sink:
topics.append(sink['recognized_object_array_topic'])
# subscribe to the output of the detection pipeline
for topic in topics:
rospy.Subscriber(topic, RecognizedObjectArray,
self.callback_recognized_object_array)
rospy.loginfo('Subscribed to the ' + topic + ' topic.')
# look for a cell that contains a cropper to select the ROI
self.cropper = None
for cell in self.plasm.cells():
if 'crop_enabled' in cell.params:
self.cropper = cell
# actionlib stuff
self.server = actionlib.SimpleActionServer('recognize_objects',
ObjectRecognitionAction,
self.execute, False)
self.server.start()
rospy.loginfo('ORK server started')
def execute_detection(cmd,path):
parser = create_parser()
# add ecto options
scheduler_options(parser)
pid = os.getpid()
print("pid = ", pid)
args = parser.parse_args(['-c', cmd])
args = parser.parse_args(['--config_file', path])
# create the plasm that will run the detection
# args = parser.parse_args()
ork_params, _args = read_arguments(args)
plasm = create_plasm(ork_params)
# run the detection plasm
run_plasm(args, plasm)
print "Exit Finished"
def handles_training(req):
parser = create_parser(do_training=True)
#args = parser.parse_args()
args = parser.parse_args(['-c', req.cmd])
args = parser.parse_args(['--config_file', req.path])
ork_params, _args = read_arguments(args)
for _pipeline_id, pipeline_param in ork_params.iteritems():
pipeline_class = find_class([ pipeline_param['module'] ], pipeline_param['type'] )
if not pipeline_class:
raise RuntimeError('Invalid pipeline name: %s\nMake sure that the pipeline type is defined by a '
'TrainingPipeline class, in the name class function.' % pipeline_param['type'])
db_params_raw = pipeline_param['parameters'].get('json_db', {})
db_params = db_params_raw
object_ids = interpret_object_ids(db_params, pipeline_param.get('parameters', {}).get('json_object_ids', []))
print('Training %d objects.' % len(object_ids))
for object_id in object_ids:
print('computing object_id: %s' % object_id)
object_id = object_id.encode('ascii')
ork_params_sub = {}
ork_params_sub[_pipeline_id] = pipeline_param
ork_params_sub[_pipeline_id]['parameters']['object_id'] = object_id
plasm = create_plasm(ork_params_sub)
plasm.execute()
# Make sure the views exist
object_db = ObjectDb(db_params)
db_params = ObjectDbParameters(db_params)
if db_params.type == ObjectDbTypes.COUCHDB:
import couchdb
import json
dic = json.loads(db_params_raw)
couch = couchdb.Server(dic['root'])
init_object_databases(couch)
return training_srvResponse()
#!/usr/bin/env python
"""
This script tests whether a detection pipeline can be created given a config file.
It is not meant to be run as a test of object_recognition but as a test for and by each
pipeline independently.
"""
from ecto.opts import scheduler_options
from object_recognition_core.pipelines.plasm import create_plasm
from object_recognition_core.utils.training_detection_args import create_parser, read_arguments
if __name__ == '__main__':
# create an ORK parser
parser = create_parser()
# add ecto options
scheduler_options(parser)
# read the config file
args = parser.parse_args()
ork_params, args = read_arguments(args)
# override the database parameters
for cell_name, parameters in ork_params.items():
if 'parameters' in parameters and 'object_ids' in parameters[
'parameters']:
parameters['parameters']['object_ids'] = '[]'
# create the big plasm
plasm = create_plasm(ork_params)
#!/usr/bin/env python
"""
This script tests whether a detection pipeline can be created given a config file.
It is not meant to be run as a test of object_recognition but as a test for and by each
pipeline independently.
"""
from ecto.opts import scheduler_options
from object_recognition_core.pipelines.plasm import create_plasm
from object_recognition_core.utils.training_detection_args import create_parser, read_arguments
if __name__ == '__main__':
# create an ORK parser
parser = create_parser()
# add ecto options
scheduler_options(parser)
# read the config file
args = parser.parse_args()
ork_params, args = read_arguments(args)
# override the database parameters
for cell_name, parameters in ork_params.items():
if 'parameters' in parameters and 'object_ids' in parameters['parameters']:
parameters['parameters']['object_ids'] = '[]'
# create the big plasm
plasm = create_plasm(ork_params)