def __init__(self):
super().__init__()
if self.weights is not None:
caf_weights = []
for bone in WHOLEBODY_SKELETON:
caf_weights.append(
max(self.weights[bone[0] - 1], self.weights[bone[1] - 1]))
w_np = np.array(caf_weights)
caf_weights = list(w_np / np.sum(w_np) * len(caf_weights))
else:
caf_weights = None
cif = headmeta.Cif('cif',
'wholebodykp',
keypoints=WHOLEBODY_KEYPOINTS,
sigmas=WHOLEBODY_SIGMAS,
pose=WHOLEBODY_STANDING_POSE,
draw_skeleton=WHOLEBODY_SKELETON,
score_weights=WHOLEBODY_SCORE_WEIGHTS,
training_weights=self.weights)
caf = headmeta.Caf('caf',
'wholebodykp',
keypoints=WHOLEBODY_KEYPOINTS,
sigmas=WHOLEBODY_SIGMAS,
pose=WHOLEBODY_STANDING_POSE,
skeleton=WHOLEBODY_SKELETON,
training_weights=caf_weights)
cif.upsample_stride = self.upsample_stride
caf.upsample_stride = self.upsample_stride
self.head_metas = [cif, caf]
def __init__(self):
super().__init__()
cif = headmeta.Cif('cif',
'apollo',
keypoints=self.CAR_KEYPOINTS,
sigmas=self.CAR_SIGMAS,
pose=self.CAR_POSE,
draw_skeleton=self.CAR_SKELETON,
score_weights=self.CAR_SCORE_WEIGHTS)
caf = headmeta.Caf('caf',
'apollo',
keypoints=self.CAR_KEYPOINTS,
sigmas=self.CAR_SIGMAS,
pose=self.CAR_POSE,
skeleton=self.CAR_SKELETON)
cif.upsample_stride = self.upsample_stride
caf.upsample_stride = self.upsample_stride
self.head_metas = [cif, caf]