本文整理汇总了Python中mtcnn.mtcnn.MTCNN属性的典型用法代码示例。如果您正苦于以下问题：Python mtcnn.MTCNN属性的具体用法？Python mtcnn.MTCNN怎么用？Python mtcnn.MTCNN使用的例子？那么, 这里精选的属性代码示例或许可以为您提供帮助。您也可以进一步了解该属性所在类mtcnn.mtcnn的用法示例。
在下文中一共展示了mtcnn.MTCNN属性的4个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: __init__
# 需要导入模块: from mtcnn import mtcnn [as 别名]
# 或者: from mtcnn.mtcnn import MTCNN [as 别名]
def __init__(self):
self.detector = MTCNN() 示例2: frame_set_select_extract
# 需要导入模块: from mtcnn import mtcnn [as 别名]
# 或者: from mtcnn.mtcnn import MTCNN [as 别名]
def frame_set_select_extract(self, frame_set_id, opts):
"""
This method extracts faces from each frame in a frame set.
:param int frame_set_id: The frame set ID.
:param dict opts: The dict of opts.
:rtype: int
"""
detector = MTCNN()
offset_percent = 0.2
min_confidence = 0.9
debug_ = True if 'debug' in opts else False
frame_set_path = FrameSetSubDir.path(frame_set_id)
transform_set_id = TransformSetModel().insert(self.name, frame_set_id)
transform_set_path = TransformSetSubDir.path(transform_set_id)
os.makedirs(transform_set_path)
length = int(os.environ.get('MODEL_LIST_LENGTH', '100'))
offset = 0
while True:
result = FrameModel().list(frame_set_id, length=length,
offset=offset, rejected=False)
if not result:
break
for frame_id, _, rejected in result:
self._extract_faces(frame_set_path, frame_id,
transform_set_path, transform_set_id,
detector, offset_percent, min_confidence,
debug_)
offset += length
return transform_set_id 示例3: __init__
# 需要导入模块: from mtcnn import mtcnn [as 别名]
# 或者: from mtcnn.mtcnn import MTCNN [as 别名]
def __init__(self):
self.face_detector = MTCNN()
model_path = os.path.join(
os.path.dirname(os.path.realpath(__file__)),
'trained_detectors',
'mesonet.hdf5')
self.mesonet = load_model(model_path) 示例4: _extract_faces
# 需要导入模块: from mtcnn import mtcnn [as 别名]
# 或者: from mtcnn.mtcnn import MTCNN [as 别名]
def _extract_faces(self, frame_set_path, frame_id, transform_set_path,
transform_set_id, detector, offset_percent,
min_confidence, debug_):
"""
This method extracts faces from a frame.
:param str frame_set_path:The frame set path.
:param int frame_id: The frame ID.
:param str transform_set_path: The transform set path.
:param transform_set_id: The transform set ID.
:param MTCNN detector: The detector to use to detect faces.
:param float offset_percent:
:param float min_confidence: The minimum confidence value required to
accept/reject a detected face.
:param bool debug_: True if should place markers on landmarks else
False if should not.
:rtype: None
"""
frame_path = FrameFile.path(frame_set_path, frame_id, 'jpg')
img = cv2.imread(frame_path)
img_height, img_width = img.shape[:2]
results = detector.detect_faces(img)
for r in results:
if r['confidence'] < min_confidence:
continue
x, y, width, height = r['box']
adjusted_x = int(max(0, x - (0.5 * width * offset_percent)))
adjusted_y = int(max(0, y - (0.5 * height * offset_percent)))
t = x + width + (0.5 * width * offset_percent)
adjusted_right_x = int(min(img_width, t))
t = y + height + (0.5 * height * offset_percent)
adjusted_bottom_y = int(min(img_height, t))
metadata = {'face': {k: [v[0] - adjusted_x, v[1] - adjusted_y]
for k, v in r['keypoints'].items()}}
transform_id = TransformModel().insert(transform_set_id, frame_id,
json.dumps(metadata), 0)
face_crop = img[adjusted_y:adjusted_bottom_y,
adjusted_x:adjusted_right_x]
output_path = TransformFile.path(transform_set_path, transform_id,
'jpg')
if debug_ is True:
for _, v in metadata['face'].items():
cv2.drawMarker(face_crop, tuple(v), (0, 0, 255),
markerType=cv2.MARKER_DIAMOND,
markerSize=15, thickness=2)
cv2.imwrite(output_path, face_crop,
[cv2.IMWRITE_JPEG_QUALITY, 100])
debug(f'Transform with ID {transform_id:08d} at {output_path} '
f'extracted from frame with ID {frame_id:08d} at '
f'{frame_path}', 4)