def _dataset_parser(self, image_file, label_file):
objects = []
tree = ET.parse(label_file)
root = tree.getroot()
objects_handle = root.find('objects')
for single_object in objects_handle.findall('object'):
points = single_object.find('points')
object_struct = {}
pointobb = []
for point in points[:-1]:
coords = [float(coord) for coord in point.text.split(',')]
pointobb += coords
bbox = aitool.pointobb2bbox(pointobb)
bbox = aitool.xyxy2xywh(bbox)
object_struct['area'] = bbox[2] * bbox[3]
object_struct['segmentation'] = [pointobb]
object_struct['pointobb'] = pointobb
object_struct['bbox'] = bbox
object_struct['category_id'] = classes[single_object.find('possibleresult').find('name').text]
objects.append(object_struct)
return objects
def _convert_items(self, result, img_size=(1024, 1024)):
"""convert the result (single image) in pkl file to specific format (default: Faster R-CNN, bbox)
Args:
result (tuple): detection result of single image
img_size (tuple): image size (height, width)
Return:
list: converted objects
"""
objects = []
for label in range(len(result)):
bboxes = result[label]
for i in range(bboxes.shape[0]):
data = dict()
data['bbox'] = aitool.xyxy2xywh(bboxes[i][:4])
if data['bbox'][2] * data['bbox'][3] < self.min_area:
continue
data['score'] = float(bboxes[i][4])
if data['score'] < self.score_threshold:
continue
data['category_id'] = self.cat_ids[label]
objects.append(data)
return objects
def xml_parser_rovoc(label_file):
objects = []
tree = ET.parse(label_file)
root = tree.getroot()
for single_object in root.findall('object'):
robndbox = single_object.find('robndbox')
data = {}
cx = float(robndbox.find('cx').text)
cy = float(robndbox.find('cy').text)
w = float(robndbox.find('w').text)
h = float(robndbox.find('h').text)
theta = float(robndbox.find('angle').text)
thetaobb = [cx, cy, w, h, theta]
pointobb = aitool.thetaobb2pointobb(thetaobb)
bbox = aitool.thetaobb2bbox(thetaobb)
bbox = aitool.xyxy2xywh(bbox)
data['area'] = bbox[2] * bbox[3]
data['segmentation'] = [pointobb]
data['pointobb'] = pointobb
data['thetaobb'] = thetaobb
data['bbox'] = bbox
data['category_id'] = 1
objects.append(data)
return objects
def _convert_items(self, result, img_size=(1024, 1024)):
"""convert the result (single image) in pkl file to specific format (Mask OBB, bbox + maskobb)
Args:
result (tuple): detection result of single image
img_size (tuple): image size (height, width)
Return:
list: converted objects
"""
objects = []
det, seg = result
for label in range(len(det)):
if label > len(seg) - 1:
continue
bboxes = det[label]
if isinstance(seg, tuple):
segms = seg[0][label]
else:
segms = seg[label]
for i in range(bboxes.shape[0]):
data = dict()
data['bbox'] = aitool.xyxy2xywh(bboxes[i][:4])
data['score'] = float(bboxes[i][4])
if data['score'] < self.score_threshold:
continue
data['category_id'] = self.cat_ids[label]
if isinstance(segms[i]['counts'], bytes):
segms[i]['counts'] = segms[i]['counts'].decode()
data['segmentation'] = segms[i]
thetaobb, pointobb = aitool.segm2rbbox(segms[i])
data['pointobb'] = pointobb
data['thetaobb'] = thetaobb
if thetaobb[2] * thetaobb[3] < self.min_area:
continue
if not self.keep_boundary:
cx_flag = thetaobb[0] < 0 or thetaobb[0] > img_size[1] - 1
cy_flag = thetaobb[1] < 0 or thetaobb[1] > img_size[0] - 1
if cx_flag or cy_flag:
continue
objects.append(data)
return objects
def _convert_items(self, result, img_size=(1024, 1024)):
"""convert the result (single image) in pkl file to specific format (Mask R-CNN, bbox + mask)
Args:
result (tuple): detection result of single image
img_size (tuple): image size (height, width)
Return:
list: converted objects
"""
objects = []
det, seg = result
for label in range(len(det)):
bboxes = det[label]
if isinstance(seg, tuple):
segms = seg[0][label]
else:
segms = seg[label]
for i in range(bboxes.shape[0]):
data = dict()
data['bbox'] = aitool.xyxy2xywh(bboxes[i][:4])
if data['bbox'][2] * data['bbox'][3] < self.min_area:
continue
data['score'] = float(bboxes[i][4])
if data['score'] < self.score_threshold:
continue
data['category_id'] = self.cat_ids[label]
if isinstance(segms[i]['counts'], bytes):
segms[i]['counts'] = segms[i]['counts'].decode()
data['segmentation'] = segms[i]
data['pointobb'] = aitool.bbox2pointobb(bboxes[i][:4])
objects.append(data)
return objects
def xml_parser_plane(label_file):
objects = []
tree = ET.parse(label_file)
root = tree.getroot()
objects_handle = root.find('objects')
if objects_handle is None:
return []
for single_object in objects_handle.findall('object'):
points = single_object.find('points')
data = {}
pointobb = []
for point in points[:-1]:
coords = [float(coord) for coord in point.text.split(',')]
pointobb += coords
bbox = aitool.pointobb2bbox(pointobb)
bbox = aitool.xyxy2xywh(bbox)
data['area'] = bbox[2] * bbox[3]
data['segmentation'] = [pointobb]
data['pointobb'] = pointobb
data['thetaobb'] = aitool.pointobb2thetaobb(pointobb)
data['bbox'] = bbox
data['category_id'] = single_object.find('possibleresult').find(
'name').text
data['class'] = single_object.find('possibleresult').find('name').text
try:
data['score'] = single_object.find('possibleresult').find(
'probability').text
except:
print("There is not probability key")
objects.append(data)
return objects