"""
Module defining several inputs for the object recognition pipeline
All source cells will have the following outputs:
- K [cv::Mat]
The camera intrinsics matrix, as a 3x3 double cv::Mat
- depth [cv::Mat]
The rescaled depth image.
- image [cv::Mat]
A cv::Mat copy.
- points3d [cv::Mat]
The 3d points, height by width (or 1 by n_points if mask) with 3 channels (x, y and z)
"""
import ecto
from ecto_image_pipeline.io.source import create_source_class
from ecto.blackbox import BlackBoxCellInfo as CellInfo
########################################################################################################################
class SourceBase(object):
"""
This is a base class for a source: you don't need to have your source cell inherit from that class but if you do,
it will make sure that its inputs/outputs fit the ORK standard (which is good if you want to interact with
the official ORK pipelines).
You need to call the BlackBox constructor in your __init__ first and then this function. Typically, your __init__ is
class Foo(ecto.BlackBox, SourceBase):
def __init__(self, *args, **kwargs):
ecto.BlackBox.__init__(self, *args, **kwargs)
SourceBase.__init__(self)
"""
def __init__(self):
"""
This ensures that the given cell exhibits the minimal interface to be
considered a source for object recognition
"""
validate_source(self)
def validate_source(cell):
"""
This ensures that the given cell exhibits the minimal interface to be
considered a source for object recognition
"""
assert(isinstance(cell, SourceBase))
outputs = dir(cell.outputs)
#all sources must produce the following
for x in ('K_depth', 'K_image', 'image', 'depth'):
if x not in outputs:
raise NotImplementedError('This cell with doc\n%s\ndoes not correctly implement the source interface. Must have an output named %s' % (cell.__doc__, x))
#type checks
for x in ('K_depth', 'K_image', 'image', 'depth'):
type_name = cell.outputs.at(x).type_name
#TODO add more explicit types.
if type_name != 'cv::Mat':
raise NotImplementedError('The cell with doc\n%s\n does not correctly implement the source interface.\n'
'Must have an output named %s, with type %s\n'
'This cells output at %s has type %s' % (cell.__doc__, x, 'cv::Mat', x, type_name))
return cell
########################################################################################################################
[docs]class OpenNI(ecto.BlackBox, SourceBase):
"""
A source for any ORK pipeline, interface with a Kinect/ASUS Xtion pro using the openni driver
"""
def __init__(self, *args, **kwargs):
from ecto_openni import FpsMode, ResolutionMode, StreamMode
for key, val_type in [ ('image_fps', FpsMode), ('depth_fps', FpsMode), ('image_mode', ResolutionMode),
('depth_mode', ResolutionMode), ('stream_mode', StreamMode) ]:
val = kwargs.get(key, None)
if isinstance(val, str):
for enum in val_type.values.values():
if val == str(enum):
kwargs[key] = enum
ecto.BlackBox.__init__(self, *args, **kwargs)
SourceBase.__init__(self)
@staticmethod
def declare_cells(p):
return {'main': CellInfo(create_source_class('image_pipeline', 'OpenNISource'))}
@staticmethod
def declare_direct_params(p):
from ecto_openni import FpsMode, ResolutionMode, StreamMode
p.declare('image_fps', "The number of frames per second for the RGB image: %s" % str(FpsMode.values.values()))
p.declare('depth_fps', "The number of frames per second for the depth image: %s" % str(FpsMode.values.values()))
p.declare('image_mode', "The resolution for the RGB image: %s" % str(ResolutionMode.values.values()))
p.declare('depth_mode', "The resolution for the depth image: %s" % str(ResolutionMode.values.values()))
p.declare('stream_mode', "The stream mode: %s" % str(StreamMode.values.values()))
@staticmethod
def declare_forwards(_p):
return ({}, {'main': 'all'}, {'main': 'all'})
def connections(self, _p):
return [self.main]
