def convert(self):
out_project = sly.Project(
os.path.join(sly.TaskPaths.RESULTS_DIR,
self.settings['res_names']['project']),
sly.OpenMode.CREATE)
progress = sly.Progress('Dataset:', len(self.src_datasets))
for ds_name, samples_paths in self.src_datasets.items():
ds = out_project.create_dataset(ds_name)
for src_img_path in samples_paths:
try:
ann_path = self.get_ann_path(src_img_path)
if all(
(os.path.isfile(x) for x in [src_img_path, ann_path])):
ann = self.get_ann(src_img_path, ann_path)
ds.add_item_file(os.path.basename(src_img_path),
src_img_path,
ann=ann)
except Exception as e:
exc_str = str(e)
sly.logger.warn(
'Input sample skipped due to error: {}'.format(
exc_str),
exc_info=True,
extra={
'exc_str': exc_str,
'dataset_name': ds_name,
'image_name': src_img_path,
})
progress.iter_done_report()
out_meta = sly.ProjectMeta(
obj_classes=sly.ObjClassCollection(self.id_to_obj_class.values()))
out_project.set_meta(out_meta)
def convert_to_nonoverlapping(meta: sly.ProjectMeta, ann: sly.Annotation) -> (sly.ProjectMeta, sly.Annotation):
common_img = np.zeros(ann.img_size, np.int32) # size is (h, w)
for idx, lbl in enumerate(ann.labels, start=1):
if need_convert(lbl.obj_class.geometry_type):
if allow_render_for_any_shape(lbl) is True:
lbl.draw(common_img, color=idx)
else:
sly.logger.warn(
"Object of class {!r} (shape: {!r}) has non spatial shape {!r}. It will not be rendered."
.format(lbl.obj_class.name,
lbl.obj_class.geometry_type.geometry_name(),
lbl.geometry.geometry_name()))
new_classes = sly.ObjClassCollection()
new_labels = []
for idx, lbl in enumerate(ann.labels, start=1):
if not need_convert(lbl.obj_class.geometry_type):
new_labels.append(lbl.clone())
else:
if allow_render_for_any_shape(lbl) is False:
continue
# @TODO: get part of the common_img for speedup
mask = common_img == idx
if np.any(mask): # figure may be entirely covered by others
g = lbl.geometry
new_bmp = sly.Bitmap(data=mask)
if new_classes.get(lbl.obj_class.name) is None:
new_classes = new_classes.add(lbl.obj_class.clone(geometry_type=sly.Bitmap))
new_lbl = lbl.clone(geometry=new_bmp, obj_class=new_classes.get(lbl.obj_class.name))
new_labels.append(new_lbl)
new_meta = meta.clone(obj_classes=new_classes)
new_ann = ann.clone(labels=new_labels)
return (new_meta, new_ann)
def create_obj_class_collection(
classes_mapping: Dict) -> sly.ObjClassCollection:
cls_list = [
sly.ObjClass(cls_name, sly.Bitmap)
for cls_name in classes_mapping.keys()
]
return sly.ObjClassCollection(cls_list)
def _determine_model_classes(self):
if 'classes' not in self.config:
# Key-value tags are ignored as a source of class labels.
img_tags = set(tag_meta.name for tag_meta in self.project.meta.img_tag_metas if
tag_meta.value_type == sly.TagValueType.NONE)
img_tags -= set(self.config['dataset_tags'].values())
train_classes = sorted(img_tags)
else:
train_classes = self.config['classes']
if 'ignore_tags' in self.config:
for tag in self.config['ignore_tags']:
if tag in train_classes:
train_classes.remove(tag)
if len(train_classes) < 2:
raise RuntimeError('Training requires at least two input classes.')
in_classification_tags_to_idx, self.classification_tags_sorted = create_classes(train_classes)
self.classification_tags_to_idx = infer_training_class_to_idx_map(self.config['weights_init_type'],
in_classification_tags_to_idx,
sly.TaskPaths.MODEL_CONFIG_PATH,
class_to_idx_config_key=self.classification_tags_to_idx_key)
self.class_title_to_idx = {}
self.out_classes = sly.ObjClassCollection()
logger.info('Determined model internal class mapping', extra={'class_mapping': self.class_title_to_idx})
logger.info('Determined model out classes', extra={'classes': self.classification_tags_sorted})
def convert(self):
out_project = sly.Project(os.path.join(sly.TaskPaths.RESULTS_DIR, self.settings['res_names']['project']),
sly.OpenMode.CREATE)
for ds_name, sample_names in self.src_datasets.items():
progress = sly.Progress('Dataset: {!r}'.format(ds_name), len(sample_names))
progress.report_every = 10 # By default progress for 18000 samples report only every 180 - too big.
ds = out_project.create_dataset(ds_name)
for name in sample_names:
img_name = name + '.jpg'
src_img_path = os.path.join(self._imgs_dir(ds_name), img_name)
inst_path = os.path.join(self._inst_dir(ds_name), name + '.png')
try:
ann = self._generate_annotation(src_img_path, inst_path)
ds.add_item_file(img_name, src_img_path, ann=ann)
except Exception as e:
exc_str = str(e)
sly.logger.warn('Input sample skipped due to error: {}'.format(exc_str), exc_info=True, extra={
'exc_str': exc_str,
'dataset_name': ds_name,
'image': src_img_path,
})
progress.iter_done_report()
sly.logger.info("Dataset '{}' samples processing is done.".format(ds_name), extra={})
out_meta = sly.ProjectMeta(obj_classes=sly.ObjClassCollection(self._class_id_to_object_class.values()))
out_project.set_meta(out_meta)
sly.logger.info("Mapillary samples processing is done.", extra={})
def prepare_meta(meta):
new_classes = []
for cls in meta.obj_classes:
cls: sly.ObjClass
new_classes.append(cls.clone(geometry_type=GET_GEOMETRY_FROM_STR("polygon")))
meta = meta.clone(obj_classes=sly.ObjClassCollection(new_classes))
return meta
def __init__(self):
self.settings = load_json_file(sly.TaskPaths.TASK_CONFIG_PATH)
self.lists_dir = os.path.join(sly.TaskPaths.DATA_DIR, 'ImageSets/Segmentation')
self.imgs_dir = os.path.join(sly.TaskPaths.DATA_DIR, 'JPEGImages')
self.segm_dir = os.path.join(sly.TaskPaths.DATA_DIR, 'SegmentationClass')
self.inst_dir = os.path.join(sly.TaskPaths.DATA_DIR, 'SegmentationObject')
self.colors_file = os.path.join(sly.TaskPaths.DATA_DIR, 'colors.txt')
self.with_instances = os.path.isdir(self.inst_dir)
sly.logger.info('Will import data {} instance info.'.format('with' if self.with_instances else 'without'))
self.obj_classes = sly.ObjClassCollection()
self._read_datasets()
self._read_colors()
def highlight_instances(meta: sly.ProjectMeta, ann: sly.Annotation) -> (sly.ProjectMeta, sly.Annotation):
new_classes = []
new_labels = []
for idx, label in enumerate(ann.labels):
new_cls = label.obj_class.clone(name=str(idx), color=sly.color.random_rgb())
new_lbl = label.clone(obj_class=new_cls)
new_classes.append(new_cls)
new_labels.append(new_lbl)
res_meta = meta.clone(obj_classes=sly.ObjClassCollection(new_classes))
res_ann = ann.clone(labels=new_labels)
return (res_meta, res_ann)
def set_project_meta(api, project_id, state):
fg_class = sly.ObjClass(state[const.FG_NAME],
GET_GEOMETRY_FROM_STR(state[const.FG_SHAPE]),
color=sly.color.hex2rgb(state[const.FG_COLOR]))
st_class = sly.ObjClass(state[const.ST_NAME],
GET_GEOMETRY_FROM_STR(state[const.ST_SHAPE]),
color=sly.color.hex2rgb(state[const.ST_COLOR]))
meta = sly.ProjectMeta(
obj_classes=sly.ObjClassCollection([fg_class, st_class]))
api.project.update_meta(
project_id,
sly.ProjectMeta().to_json()) # clear previous labels and classes
api.project.update_meta(project_id, meta.to_json())
return fg_class, st_class
def convert():
settings = load_json_file(sly.TaskPaths.SETTINGS_PATH)
out_project = sly.Project(
os.path.join(sly.TaskPaths.RESULTS_DIR,
settings['res_names']['project']), sly.OpenMode.CREATE)
classes_collection = sly.ObjClassCollection()
instance_classes, id_to_class, class_to_color = read_colors()
src_datasets = read_datasets()
skipped_count = 0
samples_count = 0
for ds_name, sample_names in src_datasets.items():
dataset = out_project.create_dataset(ds_name)
dataset_progress = sly.Progress('Dataset {!r}'.format(ds_name),
len(sample_names))
for name in sample_names:
try:
src_img_path = osp.join(images_dir(ds_name), name)
inst_path = osp.join(instances_dir(ds_name), name)
ann, classes_collection = generate_annotation(
src_img_path, inst_path, id_to_class, class_to_color,
classes_collection)
item_name = osp.splitext(name)[0]
dataset.add_item_file(item_name, src_img_path, ann)
samples_count += 1
except Exception as e:
exc_str = str(e)
sly.logger.warn(
'Input sample skipped due to error: {}'.format(exc_str),
exc_info=True,
extra={
'exc_str': exc_str,
'dataset_name': ds_name,
'image_name': name
})
skipped_count += 1
dataset_progress.iter_done_report()
sly.logger.info('Processed.',
extra={
'samples': samples_count,
'skipped': skipped_count
})
out_meta = sly.ProjectMeta(obj_classes=classes_collection)
out_project.set_meta(out_meta)
def transform_for_instance_segmentation(meta: sly.ProjectMeta, ann: sly.Annotation) -> (sly.ProjectMeta, sly.Annotation):
new_classes = {}
for obj_class in meta.obj_classes:
obj_class: sly.ObjClass
new_class = obj_class.clone(name=obj_class.name + "-mask")
new_classes[obj_class.name] = new_class
new_class_collection = sly.ObjClassCollection(list(new_classes.values()))
new_labels = []
for label in ann.labels:
obj_class = new_classes[label.obj_class.name]
new_labels.append(label.clone(obj_class=obj_class))
res_meta = meta.clone(obj_classes=new_class_collection)
res_ann = ann.clone(labels=new_labels)
return (res_meta, res_ann)
def transform_for_detection(meta: sly.ProjectMeta, ann: sly.Annotation) -> (sly.ProjectMeta, sly.Annotation):
new_classes = sly.ObjClassCollection()
new_labels = []
for label in ann.labels:
new_class = label.obj_class.clone(name=label.obj_class.name + "-bbox", geometry_type=sly.Rectangle)
if label.obj_class.geometry_type is sly.Rectangle:
new_labels.append(label.clone(obj_class=new_class))
if new_classes.get(new_class.name) is None:
new_classes = new_classes.add(new_class)
else:
bbox = label.geometry.to_bbox()
if new_classes.get(new_class.name) is None:
new_classes = new_classes.add(new_class)
new_labels.append(label.clone(bbox, new_class))
res_meta = meta.clone(obj_classes=new_classes)
res_ann = ann.clone(labels=new_labels)
return (res_meta, res_ann)
def rename_meta_and_annotations(meta: sly.ProjectMeta, ann: sly.Annotation, suffix="original"):
def _get_new_name(current_name):
return f"{current_name}-{suffix}"
new_classes = []
for obj_class in meta.obj_classes:
obj_class: sly.ObjClass
new_classes.append(obj_class.clone(name=_get_new_name(obj_class.name)))
new_meta = meta.clone(obj_classes=sly.ObjClassCollection(new_classes))
new_labels = []
for label in ann.labels:
dest_name = _get_new_name(label.obj_class.name)
dest_class = new_meta.get_obj_class(dest_name)
new_labels.append(label.clone(obj_class=dest_class))
new_ann = ann.clone(labels=new_labels)
return new_meta, new_ann
def merge(api: sly.Api, task_id, context, state, app_logger):
classes = _merge(CLASSES_INFO, META1.obj_classes, META2.obj_classes,
state["mergeClasses"], state["resolveClasses"])
tags = _merge(TAGS_INFO, META1.tag_metas, META2.tag_metas,
state["mergeTags"], state["resolveTags"])
res_meta = sly.ProjectMeta(obj_classes=sly.ObjClassCollection(classes),
tag_metas=sly.TagMetaCollection(tags),
project_type=PROJECT1.type)
res_project = api.project.create(
state["workspaceId"],
state["resultProjectName"],
type=PROJECT1.type,
description=f"{PROJECT1.name} + {PROJECT2.name}",
change_name_if_conflict=True)
api.project.update_meta(res_project.id, res_meta.to_json())
api.project.update_custom_data(
res_project.id, {
"project1": {
"id": PROJECT1.id,
"name": PROJECT1.name
},
"project2": {
"id": PROJECT2.id,
"name": PROJECT2.name
}
})
fields = [
{
"field": "data.createdProjectId",
"payload": res_project.id
},
{
"field": "data.createdProjectName",
"payload": res_project.name
},
]
api.app.set_fields(task_id, fields)
app_logger.info("Project is created",
extra={
'project_id': res_project.id,
'project_name': res_project.name
})
#api.task.set_output_project(task_id, res_project.id, res_project.name)
my_app.stop()
def upload_project_meta(api, project_id, config_yaml_info):
classes = []
for class_id, class_name in enumerate(config_yaml_info["names"]):
yaml_class_color = config_yaml_info["colors"][class_id]
obj_class = sly.ObjClass(name=class_name,
geometry_type=sly.Rectangle,
color=yaml_class_color)
classes.append(obj_class)
tags_arr = [
sly.TagMeta(name="train", value_type=sly.TagValueType.NONE),
sly.TagMeta(name="val", value_type=sly.TagValueType.NONE)
]
project_meta = sly.ProjectMeta(
obj_classes=sly.ObjClassCollection(items=classes),
tag_metas=sly.TagMetaCollection(items=tags_arr))
api.project.update_meta(project_id, project_meta.to_json())
return project_meta
def construct_model_meta(model):
names = model.module.names if hasattr(model, 'module') else model.names
colors = None
if hasattr(model, 'module') and hasattr(model.module, 'colors'):
colors = model.module.colors
elif hasattr(model, 'colors'):
colors = model.colors
else:
colors = []
for i in range(len(names)):
colors.append(sly.color.generate_rgb(exist_colors=colors))
obj_classes = [sly.ObjClass(name, sly.Rectangle, color) for name, color in zip(names, colors)]
tags = [sly.TagMeta(CONFIDENCE, sly.TagValueType.ANY_NUMBER)]
meta = sly.ProjectMeta(obj_classes=sly.ObjClassCollection(obj_classes),
tag_metas=sly.TagMetaCollection(tags))
return meta
def preview_augs(api: sly.Api, task_id, augs, infos, py_code=None):
img_info, img = get_random_image(api)
ann_json = api.annotation.download(img_info.id).annotation
ann = sly.Annotation.from_json(ann_json, meta)
res_meta, res_img, res_ann = sly.imgaug_utils.apply(augs, meta, img, ann)
file_info = save_preview_image(api, task_id, res_img)
# rename polygonal labels in existing annotation to keep them in gallery in before section
# cheat code ############################################
_labels_new_classes = []
_new_classes = {}
for label in ann.labels:
label: sly.Label
if type(label.obj_class.geometry_type) is sly.Rectangle:
new_name = f"{label.obj_class.name}_polygon_for_gallery"
if new_name not in _new_classes:
_new_classes[new_name] = label.obj_class.clone(name=new_name)
_labels_new_classes.append(label.clone(obj_class=_new_classes[new_name]))
else:
_labels_new_classes.append(label.clone())
_meta_renamed_polygons = sly.ProjectMeta(obj_classes=sly.ObjClassCollection(list(_new_classes.values())))
gallery_meta = res_meta.merge(_meta_renamed_polygons)
# cheat code ############################################
gallery, sync_keys = ui.get_gallery(project_meta=gallery_meta,
urls=[img_info.full_storage_url, file_info.full_storage_url],
card_names=["original", "augmented"],
img_labels=[_labels_new_classes, res_ann.labels])
fields = [
{"field": "data.gallery", "payload": gallery},
{"field": "state.galleryOptions.syncViewsBindings", "payload": sync_keys},
{"field": "state.previewPipelineLoading", "payload": False},
{"field": "state.previewAugLoading", "payload": False},
]
if len(infos) == 1 and py_code is None:
fields.append({"field": "state.previewPy", "payload": infos[0]["python"]})
else:
if py_code is None:
py_code = sly.imgaug_utils.pipeline_to_python(infos, random_order=False)
fields.append({"field": "state.previewPy", "payload": py_code})
api.task.set_fields(task_id, fields)
def main():
args = parse_args()
with open(args.in_file) as f:
lines = f.readlines()
names_list = [ln for ln in (line.strip() for line in lines) if ln]
out_classes = sly.ObjClassCollection(items=[
sly.ObjClass(name=name, geometry_type=sly.Rectangle)
for name in names_list
])
cls_mapping = {x: idx for idx, x in enumerate(names_list)}
res_cfg = {
SETTINGS: {},
'out_classes': out_classes.to_json(),
'class_title_to_idx': cls_mapping,
}
config_filename = os.path.join(args.out_dir,
sly.TaskPaths.MODEL_CONFIG_NAME)
dump_json_file(res_cfg, config_filename, indent=4)
print('Done: {} -> {}'.format(args.in_file, config_filename))
def transform_for_segmentation(meta: sly.ProjectMeta, ann: sly.Annotation) -> (sly.ProjectMeta, sly.Annotation):
new_classes = {}
class_masks = {}
for obj_class in meta.obj_classes:
obj_class: sly.ObjClass
new_class = obj_class.clone(name=obj_class.name + "-mask")
new_classes[obj_class.name] = new_class
class_masks[obj_class.name] = np.zeros(ann.img_size, np.uint8)
new_class_collection = sly.ObjClassCollection(list(new_classes.values()))
for label in ann.labels:
label.draw(class_masks[label.obj_class.name], color=255)
new_labels = []
for class_name, white_mask in class_masks.items():
mask = white_mask == 255
obj_class = new_classes[class_name]
bitmap = sly.Bitmap(data=mask)
new_labels.append(sly.Label(geometry=bitmap, obj_class=obj_class))
res_meta = meta.clone(obj_classes=new_class_collection)
res_ann = ann.clone(labels=new_labels)
return (res_meta, res_ann)
# coding: utf-8
import os, cv2
import numpy as np
import supervisely_lib as sly
from supervisely_lib.io.json import load_json_file
classes_dict = sly.ObjClassCollection()
default_classes_colors = {
(0, 0, 0): 'unknown',
(255, 0, 0): 'window',
(255, 128, 0): 'door',
(0, 255, 0): 'ground_floor',
(255, 255, 0): 'facade',
(128, 255, 255): 'sky',
(0, 0, 255): 'roof',
(128, 0, 255): 'balkony'
}
def read_datasets(all_ann):
src_datasets = {}
if not os.path.isdir(all_ann):
raise RuntimeError(
'There is no directory {}, but it is necessary'.format(all_ann))
sample_names = []
for file in os.listdir(all_ann):
if file.endswith('.png'):
sample_names.append(os.path.splitext(file)[0])
src_datasets['dataset'] = sample_names
sly.logger.info('Found source dataset with {} sample(s).'.format(
def __init__(self):
self.settings = json.load(open(sly.TaskPaths.TASK_CONFIG_PATH))
self.colors_file = os.path.join(sly.TaskPaths.DATA_DIR, 'config.json')
self.obj_classes = sly.ObjClassCollection()
self._read_colors()
self._read_datasets()
src_project = sly.Project(directory=src_project_dir, mode=sly.OpenMode.READ)
dst_project_dir = os.path.join(sly.TaskPaths.OUT_PROJECTS_DIR, dst_project_name)
dst_project = sly.Project(directory=dst_project_dir, mode=sly.OpenMode.CREATE)
tag_meta_train = sly.TagMeta(train_tag_name, sly.TagValueType.NONE)
tag_meta_val = sly.TagMeta(val_tag_name, sly.TagValueType.NONE)
bbox_class_mapping = {
obj_class.name: (
obj_class if (obj_class.geometry_type == sly.Rectangle)
else sly.ObjClass(obj_class.name + '_bbox', sly.Rectangle, color=obj_class.color))
for obj_class in src_project.meta.obj_classes}
dst_meta = src_project.meta.clone(
obj_classes=sly.ObjClassCollection(bbox_class_mapping.values()),
tag_metas=src_project.meta.tag_metas.add_items([tag_meta_train, tag_meta_val]))
dst_project.set_meta(dst_meta)
crop_side_fraction = (min_crop_side_fraction, max_crop_side_fraction)
total_images = api.project.get_images_count(src_project_info.id)
if total_images <= 1:
raise RuntimeError('Need at least 2 images in a project to prepare a training set (at least 1 each for training '
'and validation).')
is_train_image = sly_dataset.partition_train_val(total_images, validation_fraction)
# Iterate over datasets and items.
image_idx = 0
for dataset in src_project:
sly.logger.info('Dataset processing', extra={'dataset_name': dataset.name})
for ds_name, img_paths in zip(datasets, dataset_images):
ds = api.dataset.create(project.id, ds_name)
print('Dataset {!r} has been sucessfully creates: id={}'.format(ds.name, ds.id))
for img_path in img_paths:
img_hash = api.image.upload_path(img_path)
image_info = api.image.add(ds.id, sly.fs.get_file_name(img_path), img_hash)
print('Image (id={}, name={}) has been sucessfully added'.format(image_info.id, image_info.name))
print("Number of images in created projects: ", api.project.get_images_count(project.id))
#define object classes
class_person = sly.ObjClass('person', sly.Rectangle, color=[255, 0, 0])
class_car = sly.ObjClass('car', sly.Polygon, color=[0, 255, 0])
class_road = sly.ObjClass('road', sly.Bitmap, color=[0, 0, 255])
obj_class_collection = sly.ObjClassCollection([class_person, class_car, class_road])
#define tags for images
tagmeta_weather = sly.TagMeta(name='weather',
value_type=sly.TagValueType.ONEOF_STRING,
possible_values=['rain', 'sun', 'cloud'],
color=[153, 0, 153])
tagmeta_annotate = sly.TagMeta('to_annotation', sly.TagValueType.NONE)
#define tags for objects
tagmeta_vehicle_type = sly.TagMeta('vehicle_type', sly.TagValueType.ONEOF_STRING, ['sedan', 'suv', 'hatchback'])
tagmeta_confidence = sly.TagMeta('confidence', sly.TagValueType.ANY_NUMBER)
tagmeta_collection = sly.TagMetaCollection(
[tagmeta_weather, tagmeta_annotate, tagmeta_vehicle_type, tagmeta_confidence])
def __init__(self):
self.settings = json.load(open(sly.TaskPaths.TASK_CONFIG_PATH))
self.obj_classes = sly.ObjClassCollection()
self.tag_metas = sly.TagMetaCollection()
def main():
api = sly.Api.from_env()
# read source project
src_project = api.project.get_info_by_id(PROJECT_ID)
if src_project.type != str(sly.ProjectType.IMAGES):
raise RuntimeError("Project {!r} has type {!r}. App works only with type {!r}"
.format(src_project.name, src_project.type, sly.ProjectType.IMAGES))
src_project_meta_json = api.project.get_meta(src_project.id)
src_project_meta = sly.ProjectMeta.from_json(src_project_meta_json)
# create destination project
DST_PROJECT_NAME = "{} (rasterized)".format(src_project.name)
dst_project = api.project.create(WORKSPACE_ID, DST_PROJECT_NAME, description="rasterized", change_name_if_conflict=True)
sly.logger.info('Destination project is created.', extra={'project_id': dst_project.id, 'project_name': dst_project.name})
# mapping polygons -> bitmaps
new_classes_lst = []
for cls in src_project_meta.obj_classes:
if need_convert(cls.geometry_type):
new_class = cls.clone(geometry_type=sly.Bitmap)
else:
new_class = cls.clone()
new_classes_lst.append(new_class)
dst_classes = sly.ObjClassCollection(new_classes_lst)
# create destination meta
dst_project_meta = src_project_meta.clone(obj_classes=dst_classes)
api.project.update_meta(dst_project.id, dst_project_meta.to_json())
def convert_to_nonoverlapping(src_ann: sly.Annotation) -> sly.Annotation:
common_img = np.zeros(src_ann.img_size, np.int32) # size is (h, w)
for idx, lbl in enumerate(src_ann.labels, start=1):
if need_convert(lbl.obj_class.geometry_type):
if allow_render_non_spatial_for_any_shape(lbl) == True:
lbl.draw(common_img, color=idx)
else:
sly.logger.warn(
"Object of class {!r} (class shape: {!r}) has non spatial shape {!r}. It will not be rendered."
.format(lbl.obj_class.name,
lbl.obj_class.geometry_type.geometry_name(),
lbl.geometry.geometry_name()))
new_labels = []
for idx, lbl in enumerate(src_ann.labels, start=1):
new_cls = dst_project_meta.obj_classes.get(lbl.obj_class.name)
if not need_convert(lbl.obj_class.geometry_type):
new_lbl = lbl.clone(obj_class=new_cls)
new_labels.append(new_lbl)
else:
if allow_render_non_spatial_for_any_shape(lbl) == False:
continue
mask = common_img == idx
if np.any(mask): # figure may be entirely covered by others
g = lbl.geometry
new_bmp = sly.Bitmap(data=mask,
labeler_login=g.labeler_login,
updated_at=g.updated_at,
created_at=g.created_at)
new_lbl = lbl.clone(geometry=new_bmp, obj_class=new_cls)
new_labels.append(new_lbl)
return src_ann.clone(labels=new_labels)
for ds_info in api.dataset.get_list(src_project.id):
ds_progress = sly.Progress('Processing dataset: {!r}/{!r}'.format(src_project.name, ds_info.name),
total_cnt=ds_info.images_count)
dst_dataset = api.dataset.create(dst_project.id, ds_info.name)
img_infos_all = api.image.get_list(ds_info.id)
for img_infos in sly.batched(img_infos_all):
img_names, img_ids, img_metas = zip(*((x.name, x.id, x.meta) for x in img_infos))
ann_infos = api.annotation.download_batch(ds_info.id, img_ids)
anns = [sly.Annotation.from_json(x.annotation, src_project_meta) for x in ann_infos]
new_anns = [convert_to_nonoverlapping(ann) for ann in anns]
new_img_infos = api.image.upload_ids(dst_dataset.id, img_names, img_ids, metas=img_metas)
new_img_ids = [x.id for x in new_img_infos]
api.annotation.upload_anns(new_img_ids, new_anns)
ds_progress.iters_done_report(len(img_infos))
api.task.set_output_project(task_id, dst_project.id, dst_project.name)
def synthesize(api: sly.Api,
task_id,
state,
meta: sly.ProjectMeta,
image_infos,
labels,
bg_images,
cache_dir,
preview=True):
progress_cb = refresh_progress_preview
if preview is False:
progress_cb = refresh_progress
augs = yaml.safe_load(state["augs"])
sly.logger.info("Init augs from yaml file")
aug.init_fg_augs(augs)
visibility_threshold = augs['objects'].get('visibility', 0.8)
classes = state["selectedClasses"]
bg_info = random.choice(bg_images)
sly.logger.info("Download background")
bg = api.image.download_np(bg_info.id)
sly.logger.debug(f"BG shape: {bg.shape}")
res_image = bg.copy()
res_labels = []
# sequence of objects that will be generated
res_classes = []
to_generate = []
for class_name in classes:
original_class: sly.ObjClass = meta.get_obj_class(class_name)
res_classes.append(original_class.clone(geometry_type=sly.Bitmap))
count_range = augs["objects"]["count"]
count = random.randint(*count_range)
for i in range(count):
to_generate.append(class_name)
random.shuffle(to_generate)
res_meta = sly.ProjectMeta(obj_classes=sly.ObjClassCollection(res_classes))
progress = sly.Progress("Processing foregrounds", len(to_generate))
progress_cb(api, task_id, progress)
progress_every = max(10, int(len(to_generate) / 20))
cover_img = np.zeros(res_image.shape[:2], np.int32) # size is (h, w)
objects_area = defaultdict(lambda: defaultdict(float))
cached_images = {}
# generate objects
for idx, class_name in enumerate(to_generate, start=1):
if class_name not in labels:
progress.iter_done_report()
continue
image_id = random.choice(list(labels[class_name].keys()))
label: sly.Label = random.choice(labels[class_name][image_id])
if image_id in cached_images:
source_image = cached_images[image_id]
else:
image_info = image_infos[image_id]
source_image = _get_image_using_cache(api, cache_dir, image_id,
image_info)
cached_images[image_id] = source_image
label_img, label_mask = get_label_foreground(source_image, label)
#sly.image.write(os.path.join(cache_dir, f"{index}_label_img.png"), label_img)
#sly.image.write(os.path.join(cache_dir, f"{index}_label_mask.png"), label_mask)
label_img, label_mask = aug.apply_to_foreground(label_img, label_mask)
#sly.image.write(os.path.join(cache_dir, f"{index}_aug_label_img.png"), label_img)
#sly.image.write(os.path.join(cache_dir, f"{index}_aug_label_mask.png"), label_mask)
label_img, label_mask = aug.resize_foreground_to_fit_into_image(
res_image, label_img, label_mask)
#label_area = g.area
find_place = False
for attempt in range(3):
origin = aug.find_origin(res_image.shape, label_mask.shape)
g = sly.Bitmap(label_mask[:, :, 0].astype(bool),
origin=sly.PointLocation(row=origin[1],
col=origin[0]))
difference = count_visibility(cover_img, g, idx, origin[0],
origin[1])
allow_placement = True
for object_idx, diff in difference.items():
new_area = objects_area[object_idx]['current'] - diff
visibility_portion = new_area / objects_area[object_idx][
'original']
if visibility_portion < visibility_threshold:
#sly.logger.warn(f"Object '{idx}', attempt {attempt + 1}: "
# f"visible portion ({visibility_portion}) < threshold ({visibility_threshold})")
allow_placement = False
break
if allow_placement is True:
find_place = True
break
else:
continue
if find_place is False:
sly.logger.warn(
f"Object '{idx}' is skipped: can not be placed to satisfy visibility threshold"
)
continue
try:
aug.place_fg_to_bg(label_img, label_mask, res_image, origin[0],
origin[1])
g.draw(cover_img, color=idx)
for object_idx, diff in difference.items():
objects_area[object_idx]['current'] -= diff
current_obj_area = g.area
objects_area[idx]['current'] = current_obj_area
objects_area[idx]['original'] = current_obj_area
res_labels.append(sly.Label(g, res_meta.get_obj_class(class_name)))
except Exception as e:
#sly.logger.warn(repr(e))
sly.logger.warn(
f"FG placement error:: label shape: {label_img.shape}; mask shape: {label_mask.shape}",
extra={"error": repr(e)})
progress.iter_done_report()
if idx % progress_every == 0: # progress.need_report():
progress_cb(api, task_id, progress)
progress_cb(api, task_id, progress)
res_ann = sly.Annotation(img_size=bg.shape[:2], labels=res_labels)
# debug visualization
# sly.image.write(os.path.join(cache_dir, "__res_img.png"), res_image)
#res_ann.draw(res_image)
#sly.image.write(os.path.join(cache_dir, "__res_ann.png"), res_image)
res_meta, res_ann = rasterize.convert_to_nonoverlapping(res_meta, res_ann)
return res_image, res_ann, res_meta
def generate(api: sly.Api, task_id, context, state, app_logger):
global PRODUCT_TAGS
products_count = len(PRODUCTS.keys())
train_count = state["trainCount"]
val_count = state["valCount"]
total_count = products_count * (train_count + val_count)
augs_settings = yaml.safe_load(state["augs"])
augs.init_fg_augs(augs_settings)
PRODUCT_TAGS = PRODUCT_TAGS.add_items([TRAIN_TAG, VAL_TAG])
res_meta = sly.ProjectMeta(
obj_classes=sly.ObjClassCollection([RESULT_CLASS]),
tag_metas=PRODUCT_TAGS
)
res_project = api.project.create(WORKSPACE_ID, state["outputProjectName"], change_name_if_conflict=True)
api.project.update_meta(res_project.id, res_meta.to_json())
progress = sly.Progress("Generating", total_count)
for product_id in PRODUCTS.keys():
dataset = api.dataset.create(res_project.id, str(product_id))
tag_meta = PRODUCT_TAGS.get(product_id)
if tag_meta is None:
raise ValueError(f"TagMeta {product_id} not found")
# cache images for one product
images = {}
for image_id in PRODUCTS[product_id].keys():
images[image_id] = sly.image.read(IMAGE_PATH[image_id])
name_index = 0
for batch in sly.batched([TRAIN_TAG] * train_count + [VAL_TAG] * val_count, batch_size=10):
final_images = []
final_anns = []
final_names = []
for tag in batch:
image_id = random.choice(list(PRODUCTS[product_id].keys()))
img = images[image_id]
ann = random.choice(list(PRODUCTS[product_id][image_id]))
label_image = None
label_mask = None
label_preview = None
retry_count = 5
for retry_idx in range(5):
try:
label_image, label_mask, label_preview = \
try_generate_example(
augs_settings,
augs,
preview=True,
product_id=product_id,
img=img,
ann=ann
)
break
except Exception as e:
if retry_idx == retry_count - 1:
raise e
continue
res_ann = sly.Annotation(label_image.shape[:2],
labels=[label_preview],
img_tags=sly.TagCollection([tag, sly.Tag(tag_meta)]))
final_images.append(label_image)
final_anns.append(res_ann)
final_names.append("{:05d}.jpg".format(name_index))
name_index += 1
new_images = api.image.upload_nps(dataset.id, final_names, final_images)
new_image_ids = [image_info.id for image_info in new_images]
api.annotation.upload_anns(new_image_ids, final_anns)
progress.iters_done_report(len(batch))
refresh_progress(api, task_id, progress)
refresh_progress(api, task_id, progress)
res_project = api.project.get_info_by_id(res_project.id)
fields = [
{"field": "data.started", "payload": False},
{"field": "data.resProjectId", "payload": res_project.id},
{"field": "data.resProjectName", "payload": res_project.name},
{"field": "data.resProjectPreviewUrl",
"payload": api.image.preview_url(res_project.reference_image_url, 100, 100)},
]
api.task.set_fields(task_id, fields)
api.task.set_output_project(task_id, res_project.id, res_project.name)
app.stop()
def convert(api: sly.Api, task_id, context, state, app_logger):
api.task.set_field(task_id, "data.started", True)
TEAM_ID = int(os.environ['context.teamId'])
WORKSPACE_ID = int(os.environ['context.workspaceId'])
PROJECT_ID = int(os.environ['modal.state.slyProjectId'])
src_project = api.project.get_info_by_id(PROJECT_ID)
if src_project.type != str(sly.ProjectType.IMAGES):
raise RuntimeError(
"Project {!r} has type {!r}. App works only with type {!r}".format(
src_project.name, src_project.type, sly.ProjectType.IMAGES))
src_meta_json = api.project.get_meta(src_project.id)
src_meta = sly.ProjectMeta.from_json(src_meta_json)
new_classes = []
need_action = False
selectors = state["selectors"]
for cls in src_meta.obj_classes:
cls: sly.ObjClass
dest = selectors[cls.name]
if dest == REMAIN_UNCHANGED:
new_classes.append(cls)
else:
need_action = True
new_classes.append(
cls.clone(geometry_type=GET_GEOMETRY_FROM_STR(dest)))
if need_action is False:
fields = [{
"field": "state.showWarningDialog",
"payload": True
}, {
"field": "data.started",
"payload": False,
}]
api.task.set_fields(task_id, fields)
return
dst_project = api.project.create(src_project.workspace_id,
src_project.name + "(new shapes)",
description="new shapes",
change_name_if_conflict=True)
sly.logger.info('Destination project is created.',
extra={
'project_id': dst_project.id,
'project_name': dst_project.name
})
dst_meta = src_meta.clone(obj_classes=sly.ObjClassCollection(new_classes))
api.project.update_meta(dst_project.id, dst_meta.to_json())
total_progress = api.project.get_images_count(src_project.id)
current_progress = 0
ds_progress = sly.Progress('Processing:', total_cnt=total_progress)
for ds_info in api.dataset.get_list(src_project.id):
dst_dataset = api.dataset.create(dst_project.id, ds_info.name)
img_infos_all = api.image.get_list(ds_info.id)
for img_infos in sly.batched(img_infos_all):
img_names, img_ids, img_metas = zip(*((x.name, x.id, x.meta)
for x in img_infos))
ann_infos = api.annotation.download_batch(ds_info.id, img_ids)
anns = [
sly.Annotation.from_json(x.annotation, src_meta)
for x in ann_infos
]
new_anns = [convert_annotation(ann, dst_meta) for ann in anns]
new_img_infos = api.image.upload_ids(dst_dataset.id,
img_names,
img_ids,
metas=img_metas)
new_img_ids = [x.id for x in new_img_infos]
api.annotation.upload_anns(new_img_ids, new_anns)
current_progress += len(img_infos)
api.task.set_field(task_id, "data.progress",
int(current_progress * 100 / total_progress))
ds_progress.iters_done_report(len(img_infos))
api.task.set_output_project(task_id, dst_project.id, dst_project.name)
# to get correct "reference_image_url"
res_project = api.project.get_info_by_id(dst_project.id)
fields = [{
"field": "data.resultProject",
"payload": dst_project.name,
}, {
"field": "data.resultProjectId",
"payload": dst_project.id,
}, {
"field":
"data.resultProjectPreviewUrl",
"payload":
api.image.preview_url(res_project.reference_image_url, 100, 100),
}]
api.task.set_fields(task_id, fields)
my_app.stop()
def do(**kwargs):
api = sly.Api.from_env()
src_project = api.project.get_info_by_id(PROJECT_ID)
if src_project.type != str(sly.ProjectType.IMAGES):
raise Exception(
"Project {!r} has type {!r}. App works only with type {!r}".format(
src_project.name, src_project.type, sly.ProjectType.IMAGES))
src_project_meta_json = api.project.get_meta(src_project.id)
src_project_meta = sly.ProjectMeta.from_json(src_project_meta_json)
# check that project has anyshape classes
find_anyshape = False
new_classes_lst = []
for cls in src_project_meta.obj_classes:
if cls.geometry_type == sly.AnyGeometry:
find_anyshape = True
continue
new_classes_lst.append(cls.clone())
dst_classes = sly.ObjClassCollection(new_classes_lst)
if find_anyshape is False:
raise Exception(
"Project {!r} doesn't have classes with shape \"Any\"".format(
src_project.name))
# create destination project
dst_name = src_project.name if _SUFFIX in src_project.name else src_project.name + _SUFFIX
dst_project = api.project.create(WORKSPACE_ID,
dst_name,
description=_SUFFIX,
change_name_if_conflict=True)
sly.logger.info('Destination project is created.',
extra={
'project_id': dst_project.id,
'project_name': dst_project.name
})
dst_project_meta = src_project_meta.clone(obj_classes=dst_classes)
api.project.update_meta(dst_project.id, dst_project_meta.to_json())
def convert_annotation(src_ann, dst_project_meta):
new_labels = []
for idx, lbl in enumerate(src_ann.labels):
lbl: sly.Label
if lbl.obj_class.geometry_type == sly.AnyGeometry:
actual_geometry = type(lbl.geometry)
new_class_name = "{}_{}".format(
lbl.obj_class.name, actual_geometry.geometry_name())
new_class = dst_project_meta.get_obj_class(new_class_name)
if new_class is None:
new_class = sly.ObjClass(name=new_class_name,
geometry_type=actual_geometry,
color=sly.color.random_rgb())
dst_project_meta = dst_project_meta.add_obj_class(
new_class)
api.project.update_meta(dst_project.id,
dst_project_meta.to_json())
new_labels.append(lbl.clone(obj_class=new_class))
else:
new_labels.append(lbl)
return src_ann.clone(labels=new_labels), dst_project_meta
for ds_info in api.dataset.get_list(src_project.id):
ds_progress = sly.Progress('Dataset: {!r}'.format(ds_info.name),
total_cnt=ds_info.images_count)
dst_dataset = api.dataset.create(dst_project.id, ds_info.name)
img_infos_all = api.image.get_list(ds_info.id)
for img_infos in sly.batched(img_infos_all):
img_names, img_ids, img_metas = zip(*((x.name, x.id, x.meta)
for x in img_infos))
ann_infos = api.annotation.download_batch(ds_info.id, img_ids)
anns = [
sly.Annotation.from_json(x.annotation, src_project_meta)
for x in ann_infos
]
new_anns = []
for ann in anns:
new_ann, dst_project_meta = convert_annotation(
ann, dst_project_meta)
new_anns.append(new_ann)
new_img_infos = api.image.upload_ids(dst_dataset.id,
img_names,
img_ids,
metas=img_metas)
new_img_ids = [x.id for x in new_img_infos]
api.annotation.upload_anns(new_img_ids, new_anns)
ds_progress.iters_done_report(len(img_infos))
api.task.set_output_project(task_id, dst_project.id, dst_project.name)
my_app.stop()
def import_cityscapes(api: sly.Api, task_id, context, state, app_logger):
tag_metas = sly.TagMetaCollection()
obj_classes = sly.ObjClassCollection()
dataset_names = []
storage_dir = my_app.data_dir
if INPUT_DIR:
cur_files_path = INPUT_DIR
extract_dir = os.path.join(
storage_dir,
str(Path(cur_files_path).parent).lstrip("/"))
input_dir = os.path.join(extract_dir, Path(cur_files_path).name)
archive_path = os.path.join(
storage_dir,
cur_files_path + ".tar") # cur_files_path.split("/")[-2] + ".tar"
project_name = Path(cur_files_path).name
else:
cur_files_path = INPUT_FILE
extract_dir = os.path.join(storage_dir, get_file_name(cur_files_path))
archive_path = os.path.join(storage_dir,
get_file_name_with_ext(cur_files_path))
project_name = get_file_name(INPUT_FILE)
input_dir = os.path.join(storage_dir,
get_file_name(cur_files_path)) # extract_dir
api.file.download(TEAM_ID, cur_files_path, archive_path)
if tarfile.is_tarfile(archive_path):
with tarfile.open(archive_path) as archive:
archive.extractall(extract_dir)
else:
raise Exception("No such file".format(INPUT_FILE))
new_project = api.project.create(WORKSPACE_ID,
project_name,
change_name_if_conflict=True)
tags_template = os.path.join(input_dir, "gtFine", "*")
tags_paths = glob.glob(tags_template)
tags = [os.path.basename(tag_path) for tag_path in tags_paths]
if train_tag in tags and val_tag not in tags:
split_train = True
elif trainval_tag in tags and val_tag not in tags:
split_train = True
else:
split_train = False
search_fine = os.path.join(input_dir, "gtFine", "*", "*",
"*_gt*_polygons.json")
files_fine = glob.glob(search_fine)
files_fine.sort()
search_imgs = os.path.join(input_dir, "leftImg8bit", "*", "*",
"*_leftImg8bit" + IMAGE_EXT)
files_imgs = glob.glob(search_imgs)
files_imgs.sort()
if len(files_fine) == 0 or len(files_imgs) == 0:
raise Exception('Input cityscapes format not correct')
samples_count = len(files_fine)
progress = sly.Progress('Importing images', samples_count)
images_pathes_for_compare = []
images_pathes = {}
images_names = {}
anns_data = {}
ds_name_to_id = {}
if samples_count > 2:
random_train_indexes = get_split_idxs(samples_count, samplePercent)
for idx, orig_ann_path in enumerate(files_fine):
parent_dir, json_filename = os.path.split(
os.path.abspath(orig_ann_path))
dataset_name = os.path.basename(parent_dir)
if dataset_name not in dataset_names:
dataset_names.append(dataset_name)
ds = api.dataset.create(new_project.id,
dataset_name,
change_name_if_conflict=True)
ds_name_to_id[dataset_name] = ds.id
images_pathes[dataset_name] = []
images_names[dataset_name] = []
anns_data[dataset_name] = []
orig_img_path = json_path_to_image_path(orig_ann_path)
images_pathes_for_compare.append(orig_img_path)
if not file_exists(orig_img_path):
logger.warn(
'Image for annotation {} not found is dataset {}'.format(
orig_ann_path.split('/')[-1], dataset_name))
continue
images_pathes[dataset_name].append(orig_img_path)
images_names[dataset_name].append(
sly.io.fs.get_file_name_with_ext(orig_img_path))
tag_path = os.path.split(parent_dir)[0]
train_val_tag = os.path.basename(tag_path)
if split_train is True and samples_count > 2:
if (train_val_tag == train_tag) or (train_val_tag == trainval_tag):
if idx in random_train_indexes:
train_val_tag = train_tag
else:
train_val_tag = val_tag
# tag_meta = sly.TagMeta(train_val_tag, sly.TagValueType.NONE)
tag_meta = sly.TagMeta('split', sly.TagValueType.ANY_STRING)
if not tag_metas.has_key(tag_meta.name):
tag_metas = tag_metas.add(tag_meta)
# tag = sly.Tag(tag_meta)
tag = sly.Tag(meta=tag_meta, value=train_val_tag)
json_data = json.load(open(orig_ann_path))
ann = sly.Annotation.from_img_path(orig_img_path)
for obj in json_data['objects']:
class_name = obj['label']
if class_name == 'out of roi':
polygon = obj['polygon'][:5]
interiors = [obj['polygon'][5:]]
else:
polygon = obj['polygon']
if len(polygon) < 3:
logger.warn(
'Polygon must contain at least 3 points in ann {}, obj_class {}'
.format(orig_ann_path, class_name))
continue
interiors = []
interiors = [convert_points(interior) for interior in interiors]
polygon = sly.Polygon(convert_points(polygon), interiors)
if city_classes_to_colors.get(class_name, None):
obj_class = sly.ObjClass(
name=class_name,
geometry_type=sly.Polygon,
color=city_classes_to_colors[class_name])
else:
new_color = generate_rgb(city_colors)
city_colors.append(new_color)
obj_class = sly.ObjClass(name=class_name,
geometry_type=sly.Polygon,
color=new_color)
ann = ann.add_label(sly.Label(polygon, obj_class))
if not obj_classes.has_key(class_name):
obj_classes = obj_classes.add(obj_class)
ann = ann.add_tag(tag)
anns_data[dataset_name].append(ann)
progress.iter_done_report()
out_meta = sly.ProjectMeta(obj_classes=obj_classes, tag_metas=tag_metas)
api.project.update_meta(new_project.id, out_meta.to_json())
for ds_name, ds_id in ds_name_to_id.items():
dst_image_infos = api.image.upload_paths(ds_id, images_names[ds_name],
images_pathes[ds_name])
dst_image_ids = [img_info.id for img_info in dst_image_infos]
api.annotation.upload_anns(dst_image_ids, anns_data[ds_name])
stat_dct = {
'samples': samples_count,
'src_ann_cnt': len(files_fine),
'src_img_cnt': len(files_imgs)
}
logger.info('Found img/ann pairs.', extra=stat_dct)
images_without_anns = set(files_imgs) - set(images_pathes_for_compare)
if len(images_without_anns) > 0:
logger.warn('Found source images without corresponding annotations:')
for im_path in images_without_anns:
logger.warn('Annotation not found {}'.format(im_path))
logger.info('Found classes.',
extra={
'cnt':
len(obj_classes),
'classes':
sorted([obj_class.name for obj_class in obj_classes])
})
logger.info('Created tags.',
extra={
'cnt':
len(out_meta.tag_metas),
'tags':
sorted([tag_meta.name for tag_meta in out_meta.tag_metas])
})
my_app.stop()