def __init__(self, config: dict, in_meta: ProjectMeta,
model: SingleImageInferenceBase):
super().__init__(config, in_meta, model)
# If saving the bounding boxes on which inference was called is requested, create separate classes
# for those bounding boxes by renaming the source object classes.
self._renamer_intermediate = None
if self._config[SAVE]:
renamer_intermediate = Renamer(
add_suffix=self._config[Renamer.ADD_SUFFIX],
enabled_classes=self._config[FROM_CLASSES])
# First simply rename the matching source classes.
intermediate_renamed_classes = make_renamed_classes(
in_meta.obj_classes, renamer_intermediate, skip_missing=True)
# Next, change the geometry type for the intermediate bounding box classes to Rectangle.
intermediate_renamed_rectangle_classes = ObjClassCollection(items=[
renamed_class.clone(geometry_type=Rectangle)
for renamed_class in intermediate_renamed_classes
])
# Add the renamed Rectangle classes to the output meta and set up a class mapper.
self._out_meta = self._out_meta.add_obj_classes(
intermediate_renamed_rectangle_classes)
self._intermediate_class_mapper = RenamingObjClassMapper(
dest_obj_classes=intermediate_renamed_rectangle_classes,
renamer=renamer_intermediate)
def __init__(self, config: dict, in_meta: ProjectMeta, model: SingleImageInferenceBase):
super().__init__(config, in_meta, model)
self._renamer_intermediate = None
if self._config[SAVE]:
renamer_intermediate = Renamer(add_suffix=self._config.get(Renamer.ADD_SUFFIX, ''),
enabled_classes=self._config[FROM_CLASSES])
intermediate_renamed_classes = make_renamed_classes(in_meta.obj_classes, renamer_intermediate,
skip_missing=True)
for renamed_class in intermediate_renamed_classes:
self._out_meta = self._out_meta.add_obj_class(renamed_class.clone(geometry_type=Rectangle))
self._intermediate_class_mapper = RenamingObjClassMapper(dest_obj_classes=self._out_meta.obj_classes,
renamer=renamer_intermediate)
self._out_meta = self._out_meta.add_obj_tag_metas(self._renamed_model_img_tags)
def __init__(self, config: dict, in_meta: ProjectMeta, model: SingleImageInferenceBase):
validation_schema_path = pkg_resources.resource_filename(
__name__, 'inference_modes_schemas/{}.json'.format(self.mode_name()))
MultiTypeValidator(validation_schema_path).val(INFERENCE_MODE_CONFIG, config)
self._config = deepcopy(config)
self._out_meta = in_meta
self._model = model
model_out_meta = self._model.model_out_meta
# Renamer for model classes.
renamer_model = Renamer.from_json(config.get(MODEL_CLASSES, {}))
# Add all the applicable (passing the renamer filter) renamed model classes to the output meta.
self._out_meta = self.out_meta.add_obj_classes(
make_renamed_classes(model_out_meta.obj_classes, renamer_model, skip_missing=True))
# Make a class mapper to translate from model object space to output meta space.
# TODO store the renamed model classes separately for the mapper instead of mixing in with the input annotation
# classes.
self._model_class_mapper = RenamingObjClassMapper(dest_obj_classes=self._out_meta.obj_classes,
renamer=renamer_model)
# Renamer for model tags.
self._model_tags_renamer = Renamer.from_json(config.get(MODEL_TAGS, {}))
# Rename the model output tags, set up a mapper and add them to the output meta.
self._renamed_model_tags = make_renamed_tag_metas(
model_out_meta.tag_metas, self._model_tags_renamer, skip_missing=True)
self._model_tag_meta_mapper = RenamingTagMetaMapper(dest_tag_meta_dict=self._renamed_model_tags,
renamer=self._model_tags_renamer)
self._out_meta = self._out_meta.add_tag_metas(self._renamed_model_tags)
def __init__(self, config: dict, in_meta: ProjectMeta, model: SingleImageInferenceBase):
validation_schema_path = pkg_resources.resource_filename(
__name__, 'inference_modes_schemas/{}.json'.format(self.mode_name()))
MultiTypeValidator(validation_schema_path).val(INFERENCE_MODE_CONFIG, config)
self._config = deepcopy(config)
self._out_meta = in_meta.clone()
self._model = model
model_out_meta = self._model.model_out_meta
renamer_model = Renamer.from_json(config[MODEL_CLASSES])
for out_class in make_renamed_classes(model_out_meta.obj_classes, renamer_model, skip_missing=True):
self._out_meta = self._out_meta.add_obj_class(out_class)
self._model_class_mapper = RenamingObjClassMapper(dest_obj_classes=self._out_meta.obj_classes,
renamer=renamer_model)
self._model_img_tags_renamer = Renamer(add_suffix=config[MODEL_CLASSES][Renamer.ADD_SUFFIX])
self._renamed_model_img_tags = make_renamed_tag_metas(model_out_meta.img_tag_metas, self._model_img_tags_renamer,
skip_missing=True)
self._model_img_tag_meta_mapper = RenamingTagMetaMapper(dest_tag_meta_dict=self._renamed_model_img_tags,
renamer=self._model_img_tags_renamer)
self._out_meta = self._out_meta.merge(ProjectMeta(obj_tag_metas=model_out_meta.obj_tag_metas))
class InfModeBboxes(InferenceModeBase):
@staticmethod
def mode_name():
return 'bboxes'
@classmethod
def make_default_config(cls, model_result_suffix: str) -> dict:
config = super(InfModeBboxes, cls).make_default_config(model_result_suffix)
our_config = {
FROM_CLASSES: MATCH_ALL,
PADDING: {
LEFT: '0' + PX,
TOP: '0' + PX,
RIGHT: '0' + PX,
BOTTOM: '0' + PX,
},
SAVE: False,
Renamer.ADD_SUFFIX: '_input_bbox'
}
update_strict(config, our_config)
return config
def __init__(self, config: dict, in_meta: ProjectMeta, model: SingleImageInferenceBase):
super().__init__(config, in_meta, model)
self._renamer_intermediate = None
if self._config[SAVE]:
renamer_intermediate = Renamer(add_suffix=self._config.get(Renamer.ADD_SUFFIX, ''),
enabled_classes=self._config[FROM_CLASSES])
intermediate_renamed_classes = make_renamed_classes(in_meta.obj_classes, renamer_intermediate,
skip_missing=True)
for renamed_class in intermediate_renamed_classes:
self._out_meta = self._out_meta.add_obj_class(renamed_class.clone(geometry_type=Rectangle))
self._intermediate_class_mapper = RenamingObjClassMapper(dest_obj_classes=self._out_meta.obj_classes,
renamer=renamer_intermediate)
self._out_meta = self._out_meta.add_obj_tag_metas(self._renamed_model_img_tags)
def _do_infer_annotate(self, img: np.ndarray, ann: Annotation) -> Annotation:
result_labels = []
for src_label, roi in self._all_filtered_bbox_rois(ann, self._config[FROM_CLASSES], self._config[PADDING]):
if roi is None:
result_labels.append(src_label)
else:
roi_ann = _get_annotation_for_bbox(img, roi, self._model)
result_labels.extend(replace_labels_classes(
roi_ann.labels, self._model_class_mapper, skip_missing=True))
model_img_level_tags = make_renamed_tags(roi_ann.img_tags, self._model_img_tag_meta_mapper,
skip_missing=True)
if self._config[SAVE]:
result_labels.append(
Label(geometry=roi, obj_class=self._intermediate_class_mapper.map(src_label.obj_class),
tags=model_img_level_tags))
result_labels.append(src_label.add_tags(model_img_level_tags))
return ann.clone(labels=result_labels)
@staticmethod
def _all_filtered_bbox_rois(ann: Annotation, included_classes, crop_config: dict):
for src_label in ann.labels:
effective_roi = None
if is_name_included(src_label.obj_class.name, included_classes):
bbox = src_label.geometry.to_bbox()
roi = _make_padded_rectangle((bbox.height, bbox.width), crop_config)
maybe_effective_roi = roi.translate(drow=bbox.top, dcol=bbox.left).crop(
Rectangle.from_size(ann.img_size))
if len(maybe_effective_roi) > 0:
[effective_roi] = maybe_effective_roi
yield src_label, effective_roi
class InfModeBboxes(InferenceModeBase):
@staticmethod
def mode_name():
return 'bboxes'
@classmethod
def make_default_config(cls, model_result_suffix: str) -> dict:
config = super(InfModeBboxes,
cls).make_default_config(model_result_suffix)
our_config = {
FROM_CLASSES: MATCH_ALL,
PADDING: {
LEFT: '0' + PX,
TOP: '0' + PX,
RIGHT: '0' + PX,
BOTTOM: '0' + PX,
},
SAVE: False,
Renamer.ADD_SUFFIX: '_input_bbox'
}
update_strict(config, our_config)
return config
def __init__(self, config: dict, in_meta: ProjectMeta,
model: SingleImageInferenceBase):
super().__init__(config, in_meta, model)
# If saving the bounding boxes on which inference was called is requested, create separate classes
# for those bounding boxes by renaming the source object classes.
self._renamer_intermediate = None
if self._config[SAVE]:
renamer_intermediate = Renamer(
add_suffix=self._config[Renamer.ADD_SUFFIX],
enabled_classes=self._config[FROM_CLASSES])
# First simply rename the matching source classes.
intermediate_renamed_classes = make_renamed_classes(
in_meta.obj_classes, renamer_intermediate, skip_missing=True)
# Next, change the geometry type for the intermediate bounding box classes to Rectangle.
intermediate_renamed_rectangle_classes = ObjClassCollection(items=[
renamed_class.clone(geometry_type=Rectangle)
for renamed_class in intermediate_renamed_classes
])
# Add the renamed Rectangle classes to the output meta and set up a class mapper.
self._out_meta = self._out_meta.add_obj_classes(
intermediate_renamed_rectangle_classes)
self._intermediate_class_mapper = RenamingObjClassMapper(
dest_obj_classes=intermediate_renamed_rectangle_classes,
renamer=renamer_intermediate)
def _do_infer_annotate(self, img: np.ndarray,
ann: Annotation) -> Annotation:
result_labels = []
for src_label, roi in self._all_filtered_bbox_rois(
ann, self._config[FROM_CLASSES], self._config[PADDING]):
if roi is None:
result_labels.append(src_label)
else:
roi_ann = _get_annotation_for_bbox(img, roi, self._model)
result_labels.extend(
_replace_labels_classes(roi_ann.labels,
self._model_class_mapper,
self._model_tag_meta_mapper,
skip_missing=True))
model_img_level_tags = make_renamed_tags(
roi_ann.img_tags,
self._model_tag_meta_mapper,
skip_missing=True)
if self._config[SAVE]:
result_labels.append(
Label(geometry=roi,
obj_class=self._intermediate_class_mapper.map(
src_label.obj_class),
tags=model_img_level_tags))
# Regardless of whether we need to save intermediate bounding boxes, also put the inference result tags
# onto the original source object from which we created a bounding box.
# This is necessary for e.g. classification models to work, so that they put the classification results
# onto the original object.
result_labels.append(src_label.add_tags(model_img_level_tags))
return ann.clone(labels=result_labels)
@staticmethod
def _all_filtered_bbox_rois(ann: Annotation, included_classes,
crop_config: dict):
for src_label in ann.labels:
effective_roi = None
if is_name_included(src_label.obj_class.name, included_classes):
bbox = src_label.geometry.to_bbox()
roi = _make_padded_rectangle((bbox.height, bbox.width),
crop_config)
maybe_effective_roi = roi.translate(
drow=bbox.top,
dcol=bbox.left).crop(Rectangle.from_size(ann.img_size))
if len(maybe_effective_roi) > 0:
[effective_roi] = maybe_effective_roi
yield src_label, effective_roi