def load_model(self, cnn_feature_extractor):
if os.path.exists(cnn_feature_extractor):
app_logger.info("loading... {}".format(cnn_feature_extractor))
self.feature_extractor = pickle.load(open(cnn_feature_extractor))
app_logger.info("done")
else:
self.feature_extractor = CnnFeatureExtractor(self.use_gpu, self.model, self.model_type,
self.image_feature_dim)
pickle.dump(self.feature_extractor, open(cnn_feature_extractor, 'w'))
app_logger.info("pickle.dump finished")
def __init__(self, **kwargs):
if os.path.exists(CNN_FEATURE_EXTRACTOR):
app_logger.info("loading... {}".format(CNN_FEATURE_EXTRACTOR))
self.feature_extractor = pickle.load(open(CNN_FEATURE_EXTRACTOR))
app_logger.info("done")
else:
self.feature_extractor = CnnFeatureExtractor(use_gpu, CAFFE_MODEL, MODEL_TYPE, image_feature_dim)
pickle.dump(self.feature_extractor, open(CNN_FEATURE_EXTRACTOR, 'w'))
app_logger.info("pickle.dump finished")
self.agent_service = AgentService(BRICA_CONFIG_FILE, self.feature_extractor)
self.result_logger = ResultLogger()
class VVCComponent(brica1.Component):
image_feature_count = 1
cnn_feature_extractor = CNN_FEATURE_EXTRACTOR
model = CAFFE_MODEL
model_type = MODEL_TYPE
image_feature_dim = 256 * 6 * 6
def __init__(self, n_output=10240, n_input=1):
# image_feature_count = 1
super(VVCComponent, self).__init__()
self.use_gpu = use_gpu
self.n_output = n_output
self.n_input = n_input
def set_model(self, feature_extractor):
self.feature_extractor = feature_extractor
def load_model(self, cnn_feature_extractor):
if os.path.exists(cnn_feature_extractor):
app_logger.info("loading... {}".format(cnn_feature_extractor))
self.feature_extractor = pickle.load(open(cnn_feature_extractor))
app_logger.info("done")
else:
self.feature_extractor = CnnFeatureExtractor(self.use_gpu, self.model, self.model_type,
self.image_feature_dim)
pickle.dump(self.feature_extractor, open(cnn_feature_extractor, 'w'))
app_logger.info("pickle.dump finished")
def fire(self):
observation = self.get_in_port('Isocortex#V1-Isocortex#VVC-Input').buffer
obs_array = self.feature_extractor.feature(observation, self.image_feature_count)
self.results['Isocortex#VVC-BG-Output'] = obs_array
self.results['Isocortex#VVC-UB-Output'] = obs_array
class Root(object):
def __init__(self, **kwargs):
if os.path.exists(CNN_FEATURE_EXTRACTOR):
app_logger.info("loading... {}".format(CNN_FEATURE_EXTRACTOR))
self.feature_extractor = pickle.load(open(CNN_FEATURE_EXTRACTOR, 'rb'))
app_logger.info("done")
else:
self.feature_extractor = CnnFeatureExtractor(use_gpu, CAFFE_MODEL, MODEL_TYPE, image_feature_dim)
pickle.dump(self.feature_extractor, open(CNN_FEATURE_EXTRACTOR, 'wb'))
app_logger.info("pickle.dump finished")
self.agent_service = AgentService(BRICA_CONFIG_FILE, self.feature_extractor)
self.result_logger = ResultLogger()
@cherrypy.expose()
def flush(self, identifier):
self.agent_service.initialize(identifier)
@cherrypy.expose
def create(self, identifier):
body = cherrypy.request.body.read()
reward, observation, rotation, movement = unpack(body)
inbound_logger.info('reward: {}, depth: {}'.format(reward, observation['depth']))
feature = self.feature_extractor.feature(observation)
self.result_logger.initialize()
result = self.agent_service.create(reward, feature, identifier)
self.result_logger.add_agent(self.agent_service.agents[identifier])
outbound_logger.info('action: {}'.format(result))
return str(result)
@cherrypy.expose
def step(self, identifier):
body = cherrypy.request.body.read()
reward, observation, rotation, movement = unpack(body)
inbound_logger.info('reward: {}, depth: {}'.format(reward, observation['depth']))
result = self.agent_service.step(reward, observation, identifier)
self.result_logger.step()
outbound_logger.info('result: {}'.format(result))
return str(result)
@cherrypy.expose
def reset(self, identifier):
body = cherrypy.request.body.read()
reward, success, failure, elapsed, finished = unpack_reset(body)
inbound_logger.info('reward: {}, success: {}, failure: {}, elapsed: {}'.format(
reward, success, failure, elapsed))
result = self.agent_service.reset(reward, identifier)
self.result_logger.report(success, failure, finished)
outbound_logger.info('result: {}'.format(result))
return str(result)