The config of vovnet backbone is like this.
img_backbone=dict(
type='VoVNetCP', ###use checkpoint to save memory
spec_name='V2-99-eSE',
norm_eval=True,
frozen_stages=-1,
input_ch=3,
out_features=('stage2', 'stage3', 'stage4', 'stage5',)),
It seems not use the pretrained weights to initialize it. Is it necessary to use [FCOS3D] to initialize the weights of vovnet?
The config of vovnet backbone is like this. img_backbone=dict( type='VoVNetCP', ###use checkpoint to save memory spec_name='V2-99-eSE', norm_eval=True, frozen_stages=-1, input_ch=3, out_features=('stage2', 'stage3', 'stage4', 'stage5',)),
It seems not use the pretrained weights to initialize it. Is it necessary to use [FCOS3D] to initialize the weights of vovnet?