def test_specify_type(jpg_file_path):
with qidata.open(jpg_file_path, "w") as f:
assert (DataType.IMAGE == f.type)
f.type = DataType.IMAGE_2D
assert (DataType.IMAGE_2D == f.type)
f.type = "IMAGE_2D"
assert (DataType.IMAGE_2D == f.type)
with pytest.raises(TypeError):
f.type = DataType.AUDIO
with qidata.open(jpg_file_path, "r") as f:
assert (DataType.IMAGE_2D == f.type)
def test_specialized_qidatafile(file_name, class_, datatype, valid_locs,
invalid_locs):
with qidata.open(conftest.sandboxed(file_name), "r") as _f:
assert (isinstance(_f, class_))
assert (datatype == _f.type)
with qidata.open(conftest.sandboxed(file_name), "w") as _f:
a = metadata_objects.Property(key="prop", value="10")
_f.addAnnotation("jdoe", a, None)
for invalid_loc in invalid_locs:
with pytest.raises(Exception) as e:
_f.addAnnotation("jdoe", a, invalid_loc)
assert ('Location %s is invalid' %
str(invalid_loc) == e.value.message)
for valid_loc in valid_locs:
_f.addAnnotation("jdoe", a, valid_loc)
def process(self, i, o):
cam = QiDataImage()
with qidata.open(self.params.get("image_file").get()) as _f:
# Retrieve image, convert it if necessary, then put it in a Cv::Mat
# and in QiDataImage
if ("COLOR" == self.params.get("mode").get()
and Colorspace("BGR") != _f.raw_data.colorspace):
# convert
_tmp = _f.raw_data.render().numpy_image
colorspace = 13 # AL_code for BGR
elif ("GRAYSCALE" == self.params.get("mode").get()
and Colorspace("Gray") != _f.raw_data.colorspace):
_tmp = _f.raw_data.render().numpy_image
_tmp = cv2.cvtColor(_tmp, cv2.COLOR_BGR2GRAY)
colorspace = 0 # AL_code for Gray
else:
# no convert
_tmp = _f.raw_data.numpy_image
colorspace = _f.raw_data.colorspace.al_code
cam.data.fromarray(_tmp)
cam.colorspace = colorspace
# Register the camera's position
cam.tf.tx = _f.transform.translation.x
cam.tf.ty = _f.transform.translation.y
cam.tf.tz = _f.transform.translation.z
cam.tf.rx = _f.transform.rotation.x
cam.tf.ry = _f.transform.rotation.y
cam.tf.rz = _f.transform.rotation.z
cam.tf.rw = _f.transform.rotation.w
# Register the image's timestamp
cam.ts.seconds = _f.timestamp.seconds
cam.ts.nanoseconds = _f.timestamp.nanoseconds
# Register the calibration
cam.camera_matrix.fromarray(
numpy.array(_f.raw_data.camera_info.camera_matrix))
cam.distortion_coeffs = ecto.list_of_floats(
_f.raw_data.camera_info.distortion_coeffs)
cam.rectification_matrix.fromarray(
numpy.array(_f.raw_data.camera_info.rectification_matrix))
cam.projection_matrix.fromarray(
numpy.array(_f.raw_data.camera_info.projection_matrix))
o.get("qidata_image").set(cam)
return ecto.OK
def examineContent(self):
"""
Examine all dataset's files to infer content information.
For every supported file contained in the dataset, this function will:
- check if a DataType is already defined and infer one from the file
extension if not.
- open the file and look for present annotations
Once all files have been studied, remaining annotations will be updated
with any known status that might have been present before this function
was called.
"""
_annotation_content = dict()
self._files_type = dict()
for name in self.children:
path = os.path.join(self._folder_path, name)
with qidata.open(path, "r") as _f:
for annotator, annotations in _f.annotations.iteritems():
for annotation_type in annotations.keys():
_annotation_content[(
annotator, annotation_type
)] = QiDataSet.AnnotationStatus.PARTIAL
if not self._files_type.has_key(str(_f.type)):
self._files_type[str(_f.type)] = []
self._files_type[str(_f.type)].append(name)
for _f in self.getAllFrames():
for annotator, annotations in _f.annotations.iteritems():
for annotation_type in annotations.keys():
_annotation_content[(
annotator,
annotation_type)] = QiDataSet.AnnotationStatus.PARTIAL
# For all discovered annotation, grab the previously known status
# If an annotation had a status before but was not seen, it does
# need to be kept only if it was declared as TOTAL
for key in self._annotation_content:
if QiDataSet.AnnotationStatus.TOTAL == self._annotation_content[
key]:
_annotation_content[key] = QiDataSet.AnnotationStatus.TOTAL
self._annotation_content = _annotation_content
def openChild(self, name):
"""
Open QiDataFile contained here
:param name: Name of the file or folder to open
:type name: str
.. note::
The opening mode used to open children is the opening mode of the
QiDataSet itself
"""
path = os.path.join(self._folder_path, name)
if not name in self.children:
raise IOError("%s is not a child of the current dataset" % name)
if os.path.isfile(path):
return qidata.open(path, self.mode)
# elif os.path.isdir(path):
# return QiDataSet(path, self.mode)
else:
raise IOError("%s is neither a file nor a folder" % name)
def process(self, i, o):
cams = []
cam1 = QiDataImage()
cam2 = QiDataImage()
with qidata.open(self.params.get("image_file").get()) as _f:
# Retrieve image, convert it if necessary, then put it in a Cv::Mat
# and in QiDataImage
images = []
if hasattr(_f.raw_data, "left_image"):
images.append(_f.raw_data.left_image)
images.append(_f.raw_data.right_image)
elif hasattr(_f.raw_data, "top_image"):
images.append(_f.raw_data.top_image)
images.append(_f.raw_data.bottom_image)
else:
raise RuntimeError("Given image is not stereo")
for img in images:
cam = QiDataImage()
if ("COLOR" == self.params.get("mode").get()
and Colorspace("BGR") != _f.raw_data.colorspace):
# convert
_tmp = img.render().numpy_image
colorspace = 13 # AL_code for BGR
elif ("GRAYSCALE" == self.params.get("mode").get()
and Colorspace("Gray") != _f.raw_data.colorspace):
_tmp = img.render().numpy_image
_tmp = cv2.cvtColor(_tmp, cv2.COLOR_BGR2GRAY)
colorspace = 0 # AL_code for Gray
else:
# no convert
_tmp = img.numpy_image
colorspace = _f.raw_data.colorspace.al_code
cam.data.fromarray(_tmp)
cam.colorspace = colorspace
# Register the camera's position
cam.tf.tx = _f.transform.translation.x
cam.tf.ty = _f.transform.translation.y
cam.tf.tz = _f.transform.translation.z
cam.tf.rx = _f.transform.rotation.x
cam.tf.ry = _f.transform.rotation.y
cam.tf.rz = _f.transform.rotation.z
cam.tf.rw = _f.transform.rotation.w
# Register the image's timestamp
cam.ts.seconds = _f.timestamp.seconds
cam.ts.nanoseconds = _f.timestamp.nanoseconds
# Register the calibration
cam.camera_matrix.fromarray(
numpy.array(img.camera_info.camera_matrix))
cam.distortion_coeffs = ecto.list_of_floats(
img.camera_info.distortion_coeffs)
cam.rectification_matrix.fromarray(
numpy.array(img.camera_info.rectification_matrix))
cam.projection_matrix.fromarray(
numpy.array(img.camera_info.projection_matrix))
cams.append(cam)
o.get("qidata_image_main").set(cams[0])
o.get("qidata_image_secondary").set(cams[1])
return ecto.OK