def sample_model_bg(modelnames, bgnames, n_ex_per_model, n_ex, rand=0):
""" Samples a list of models and backgrounds for building the
dataset. The models will be repeated 'n_ex_per_model' times.
modelnames: list of model names
bgnames: list of background filenames
n_ex_per_model: number of examples per model
n_ex: total number of examples
modellist: list of 'n_ex' models
bglist: list of 'n_ex' background names
"""
# Make model list
modellist = [m for m in flatten(zip(*([modelnames] * n_ex_per_model)))]
# Make background list
rand = tools.init_rand(rand)
bgrand = rand.randint(0, n_bg, n_ex)
bglist = [bgnames[r] for r in bgrand]
# Make category list
# Model root directory
model_path = os.path.join(gt.RESOURCE_PATH, "objs")
# Get the model info that's contained in the scripts
sys.path.append(model_path)
model_categories = __import__("model_categories").MODEL_CATEGORIES
# Assemble category info in dict with {modelname: category, ...}
categories = []
for categ, names in model_categories.iteritems():
categories.extend([(name, categ) for name in names])
categorydict = dict(categories)
# The actual list
categorylist = [categorydict[model] for model in modellist]
return modellist, bglist, categorylist
def __init__(self, image, renderer=None, rand=0):
# Initialize a random object
self.rand = tools.init_rand(rand)
# Initial/null values
self.image = image
self.state0 = None
self.state = None
self.score = None
self.states = []
self.proposals = []
self._store_dtype = np.dtype([("score", ("f8", 1)),
("proposal_score", ("f8", 1)),
("fwdprob", ("f8", 1)),
("bakprob", ("f8", 1)),
("flip", ("f8", 1)),
("accepted", ("b", 1))])
self.store_info = None
if renderer is None:
# Create a renderer object
self.renderer = Renderer()
else:
# Use the input renderer object
self.renderer = renderer
self.true_features = self.get_features(self.image)
# Initialize proposal mechanism
self.init_proposer(rand=rand)
# Initialize the sampler
self.init_state()
# Parameters
self.param = {
"llik_sigma": 200.,
}
def __init__(self, image, renderer=None, rand=0):
# Initialize a random object
self.rand = tools.init_rand(rand)
# Initial/null values
self.image = image
self.state0 = None
self.state = None
self.score = None
self.states = []
self.proposals = []
self._store_dtype = np.dtype([
("score", ("f8", 1)), ("proposal_score", ("f8", 1)),
("fwdprob", ("f8", 1)), ("bakprob", ("f8", 1)),
("flip", ("f8", 1)), ("accepted", ("b", 1))])
self.store_info = None
if renderer is None:
# Create a renderer object
self.renderer = Renderer()
else:
# Use the input renderer object
self.renderer = renderer
self.true_features = self.get_features(self.image)
# Initialize proposal mechanism
self.init_proposer(rand=rand)
# Initialize the sampler
self.init_state()
# Parameters
self.param = {"llik_sigma": 200.,}
def __init__(self, rand=0):
self.rand = tools.init_rand(rand)
self.init_state_info()
def __init__(self, rand=0):
self.rand = tools.init_rand(rand)
self.init_state_info()
