def getGtInstances(groundTruthList,args):
gtInstances = {}
# if there is a global statistics json, then load it
if (os.path.isfile(args.gtInstancesFile)):
if not args.quiet:
print("Loading ground truth instances from JSON.")
with open(args.gtInstancesFile) as json_file:
gtInstances = json.load(json_file)
# otherwise create it
else:
if (not args.quiet):
print("Creating ground truth instances from png files.")
gtInstances = instances2dict(groundTruthList,not args.quiet)
writeDict2JSON(gtInstances, args.gtInstancesFile)
return gtInstances
def getGtInstances(groundTruthList, args):
gtInstances = {}
# if there is a global statistics json, then load it
if (os.path.isfile(args.gtInstancesFile)):
if not args.quiet:
print("Loading ground truth instances from JSON.")
with open(args.gtInstancesFile) as json_file:
gtInstances = json.load(json_file)
# otherwise create it
else:
if (not args.quiet):
print("Creating ground truth instances from png files.")
gtInstances = instances2dict(groundTruthList, not args.quiet)
writeDict2JSON(gtInstances, args.gtInstancesFile)
return gtInstances
def _run(self, input_folder, output_folder, FORMAT):
train_path = os.path.join(input_folder, 'gtFine_trainvaltest',
'gtFine', 'train')
assert os.path.isdir(train_path)
val_path = os.path.join(input_folder, 'gtFine_trainvaltest', 'gtFine',
'val')
assert os.path.isdir(val_path)
dataset_folders = [train_path, val_path]
dataset_names = ['train', 'val']
input_image_folders = [
os.path.join(input_folder, 'leftImg8bit_trainvaltest',
'leftImg8bit', 'train'),
os.path.join(input_folder, 'leftImg8bit_trainvaltest',
'leftImg8bit', 'val')
]
self._ensure_folder_exists_and_is_clear(output_folder)
label_names = [
label.name for label in labels
if label.hasInstances and not label.ignoreInEval
]
annotation_folder = os.path.join(output_folder, 'annotations')
self._ensure_folder_exists_and_is_clear(annotation_folder)
for input_image_folder, dataset_folder, dataset_name in zip(
input_image_folders, dataset_folders, dataset_names):
self._reset_dump()
image_folder = os.path.join(output_folder, dataset_name)
self._ensure_folder_exists_and_is_clear(image_folder)
label2name = {}
for cat_id, label_name in enumerate(label_names):
label2name[name2label[label_name].id] = label_name
self.to_dump['categories'].append({
'id': cat_id + 1,
'name': label_name
})
for city_folder in os.listdir(dataset_folder):
input_image_folder_city = os.path.join(input_image_folder,
city_folder)
assert os.path.isdir(os.path.join(dataset_folder, city_folder))
for file_name in os.listdir(
os.path.join(dataset_folder, city_folder)):
file_name = os.path.join(dataset_folder, city_folder,
file_name)
assert os.path.isfile(file_name)
if file_name.endswith('_polygons.json'):
self._parse_json(file_name)
img_inst = cv2.imread(self.img_inst_name,
cv2.IMREAD_ANYDEPTH)
instance_dict = instances2dict(
[self.img_inst_name],
verbose=False)[self.img_inst_name]
for instanceId in np.unique(img_inst):
if instanceId < 1000:
continue
label = None
for value in list(instance_dict.values()):
for list_value in value:
if list_value['instID'] == instanceId:
label = list_value['labelID']
break
assert label is not None
if label not in label2name:
continue
label_to_dump = None
for category in self.to_dump['categories']:
if category['name'] == label2name[label]:
label_to_dump = category['id']
break
assert label_to_dump is not None
mask = (img_inst == instanceId).astype(np.uint8)
_, contours, _ = cv2.findContours(
mask, cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
x_min = float("inf")
y_min = float("inf")
x_max = 0
y_max = 0
for s in contours:
x, y, w, h = cv2.boundingRect(s)
x_b = x + w
y_b = y + h
x_min = min(x_min, x)
y_min = min(y_min, y)
x_max = max(x_max, x_b)
y_max = max(y_max, y_b)
area = int(np.sum(mask == 1))
segm = [
contour.ravel().tolist()
for contour in contours
]
self.to_dump['annotations'].append({
'segmentation':
segm,
'area':
area,
'iscrowd':
0,
'image_id':
self.image_id,
'bbox':
[x_min, y_min, x_max - x_min, y_max - y_min],
'category_id':
label_to_dump,
'id':
self.annotation_id,
'ignore':
0
})
self.annotation_id += 1
self.image_id += 1
self.copy(
os.path.join(input_image_folder_city,
self.image_name), image_folder)
with open(os.path.join(annotation_folder, dataset_name + '.json'),
'w') as f:
json.dump(self.to_dump, f)
def convert_cityscapes_car_only(data_dir, out_dir):
"""Convert from cityscapes format to COCO instance seg format - polygons"""
sets = [
"gtFine_val",
"gtFine_train",
# 'gtFine_test',
# 'gtCoarse_train',
# 'gtCoarse_val',
# 'gtCoarse_train_extra'
]
ann_dirs = [
"gtFine_trainvaltest/gtFine/val",
"gtFine_trainvaltest/gtFine/train",
# 'gtFine_trainvaltest/gtFine/test',
# 'gtCoarse/train',
# 'gtCoarse/train_extra',
# 'gtCoarse/val'
]
json_name = "caronly_filtered_unlabeled_%s.json"
ends_in = "%s_polygons.json"
img_id = 0
ann_id = 0
cat_id = 1
category_dict = {}
category_instancesonly = [
"car",
]
category_dict["car"] = cat_id
for data_set, ann_dir in zip(sets, ann_dirs):
print("Starting %s" % data_set)
ann_dict = {}
images = []
annotations = []
ann_dir = os.path.join(data_dir, ann_dir)
for root, _, files in os.walk(ann_dir):
for filename in files:
if filename.endswith(ends_in % data_set.split("_")[0]):
if len(images) % 50 == 0:
print("Processed %s images, %s annotations" %
(len(images), len(annotations)))
json_ann = json.load(open(os.path.join(root, filename)))
image = {}
image["id"] = img_id
img_id += 1
image["width"] = json_ann["imgWidth"]
image["height"] = json_ann["imgHeight"]
image["file_name"] = (
filename[:-len(ends_in % data_set.split("_")[0])] +
"leftImg8bit.png")
image["seg_file_name"] = (
filename[:-len(ends_in % data_set.split("_")[0])] +
"%s_instanceIds.png" % data_set.split("_")[0])
images.append(image)
fullname = os.path.join(root, image["seg_file_name"])
objects = cs.instances2dict([fullname],
verbose=False)[fullname]
bbox = []
# x
bbox.append(0)
# y
bbox.append(0)
# w
bbox.append(int(json_ann["imgWidth"]))
# h
bbox.append(int(json_ann["imgHeight"]))
seg = []
# bbox[] is x,y,w,h
# left_top
seg.append(bbox[0])
seg.append(bbox[1])
# left_bottom
seg.append(bbox[0])
seg.append(bbox[1] + bbox[3])
# right_bottom
seg.append(bbox[0] + bbox[2])
seg.append(bbox[1] + bbox[3])
# right_top
seg.append(bbox[0] + bbox[2])
seg.append(bbox[1])
ann = {}
ann["id"] = ann_id
ann_id += 1
ann["image_id"] = image["id"]
ann["segmentation"] = [seg]
ann["category_id"] = 1
ann["iscrowd"] = 0
ann["area"] = bbox[2] * bbox[3]
ann["bbox"] = bbox
annotations.append(ann)
ann_dict["images"] = images
categories = [{
"id": category_dict[name],
"name": name
} for name in category_dict]
ann_dict["categories"] = categories
ann_dict["annotations"] = annotations
print("Num categories: %s" % len(categories))
print("Num images: %s" % len(images))
print("Num annotations: %s" % len(annotations))
print(categories)
with open(os.path.join(out_dir, json_name % data_set), "w") as outfile:
outfile.write(json.dumps(ann_dict))