def init_weights(self):
if isinstance(self.decoder_input_proj, Conv2d):
caffe2_xavier_init(self.decoder_input_proj, bias=0)
self.pixel_decoder.init_weights()
for p in self.transformer_decoder.parameters():
if p.dim() > 1:
nn.init.xavier_uniform_(p)
def init_weights(self):
caffe2_xavier_init(self.lateral_conv)
caffe2_xavier_init(self.fpn_conv)
for m in [self.lateral_norm, self.fpn_norm]:
constant_init(m, 1)
for m in self.dilated_encoder_blocks.modules():
if isinstance(m, nn.Conv2d):
normal_init(m, mean=0, std=0.01)
if is_norm(m):
constant_init(m, 1)
def init_weights(self):
for module in self.fpn.values():
if hasattr(module, 'conv_out'):
caffe2_xavier_init(module.out_conv.conv)
for modules in [
self.adapt_convs.modules(),
self.extra_downsamples.modules()
]:
for module in modules:
if isinstance(module, nn.Conv2d):
caffe2_xavier_init(module)
def init_weights(self):
# retinanet_bias_init
bias_cls = bias_init_with_prob(0.01)
normal_init(self.fcos_reg, std=0.01)
normal_init(self.fcos_centerness, std=0.01)
normal_init(self.fcos_cls, std=0.01, bias=bias_cls)
for branch in [self.cls_convs, self.reg_convs]:
for module in branch.modules():
if isinstance(module, ConvModule) \
and isinstance(module.conv, nn.Conv2d):
caffe2_xavier_init(module.conv)
def init_weights(self):
"""Initialize weights."""
for i in range(0, self.num_inputs - 2):
caffe2_xavier_init(self.lateral_convs[i].conv, bias=0)
caffe2_xavier_init(self.output_convs[i].conv, bias=0)
caffe2_xavier_init(self.mask_feature, bias=0)
caffe2_xavier_init(self.encoder_in_proj, bias=0)
caffe2_xavier_init(self.encoder_out_proj.conv, bias=0)
for p in self.encoder.parameters():
if p.dim() > 1:
nn.init.xavier_uniform_(p)
def init_weights(self):
"""Initialize the weights of module."""
super(NASFCOS_FPN, self).init_weights()
for module in self.fpn.values():
if hasattr(module, 'conv_out'):
caffe2_xavier_init(module.out_conv.conv)
for modules in [
self.adapt_convs.modules(),
self.extra_downsamples.modules()
]:
for module in modules:
if isinstance(module, nn.Conv2d):
caffe2_xavier_init(module)
def _init_hint_attn_layer(self):
self.hint_attn_layer = []
for i in range(len(self.anchor_generator.strides)):
self.hint_attn_layer.append(
nn.Sequential(nn.Conv2d(256, 256, 3, padding=1),
nn.ReLU(inplace=True),
nn.Conv2d(256, 256, 3, padding=1)))
self.hint_attn_layer[i].cuda()
self.hint_attn_layer = nn.ModuleList(self.hint_attn_layer)
from mmcv.cnn import xavier_init, caffe2_xavier_init
for i in range(len(self.anchor_generator.strides)):
for m in self.hint_attn_layer[i].modules():
if isinstance(m, nn.Conv2d):
caffe2_xavier_init(m)
def init_weights(self):
"""Initialize weights."""
for i in range(0, self.num_inputs - 2):
caffe2_xavier_init(self.lateral_convs[i].conv, bias=0)
caffe2_xavier_init(self.output_convs[i].conv, bias=0)
caffe2_xavier_init(self.mask_feature, bias=0)
caffe2_xavier_init(self.last_feat_conv, bias=0)
def test_caffe_xavier_init():
conv_module = nn.Conv2d(3, 16, 3)
caffe2_xavier_init(conv_module)
def init_weights(self):
"""Initialize the weights of module."""
for m in self.modules():
if isinstance(m, nn.Conv2d):
caffe2_xavier_init(m)
def init_weights(self):
for m in self.modules():
if isinstance(m, nn.Conv2d):
caffe2_xavier_init(m)
def init_weights(self):
for m in self.modules():
if isinstance(m, nn.Conv2d):
caffe2_xavier_init(m)
elif is_norm(m):
constant_init(m, 1.0)
