How to use the mxnet.symbol.Convolution function in mxnet

act_type='relu')
        res3b3_branch2b = mx.symbol.Convolution(name='res3b3_branch2b', data=res3b3_branch2a_relu, num_filter=128,
                                                pad=(1, 1), kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn3b3_branch2b = mx.symbol.BatchNorm(name='bn3b3_branch2b', data=res3b3_branch2b, use_global_stats=True,
                                             fix_gamma=False, eps = self.eps)
        scale3b3_branch2b = bn3b3_branch2b
        res3b3_branch2b_relu = mx.symbol.Activation(name='res3b3_branch2b_relu', data=scale3b3_branch2b,
                                                    act_type='relu')
        res3b3_branch2c = mx.symbol.Convolution(name='res3b3_branch2c', data=res3b3_branch2b_relu, num_filter=512,
                                                pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn3b3_branch2c = mx.symbol.BatchNorm(name='bn3b3_branch2c', data=res3b3_branch2c, use_global_stats=True,
                                             fix_gamma=False, eps = self.eps)
        scale3b3_branch2c = bn3b3_branch2c
        res3b3 = mx.symbol.broadcast_add(name='res3b3', *[res3b2_relu, scale3b3_branch2c])
        res3b3_relu = mx.symbol.Activation(name='res3b3_relu', data=res3b3, act_type='relu')
        res4a_branch1 = mx.symbol.Convolution(name='res4a_branch1', data=res3b3_relu, num_filter=1024, pad=(0, 0),
                                              kernel=(1, 1), stride=(2, 2), no_bias=True)
        bn4a_branch1 = mx.symbol.BatchNorm(name='bn4a_branch1', data=res4a_branch1, use_global_stats=True,
                                           fix_gamma=False, eps = self.eps)
        scale4a_branch1 = bn4a_branch1
        res4a_branch2a = mx.symbol.Convolution(name='res4a_branch2a', data=res3b3_relu, num_filter=256, pad=(0, 0),
                                               kernel=(1, 1), stride=(2, 2), no_bias=True)
        bn4a_branch2a = mx.symbol.BatchNorm(name='bn4a_branch2a', data=res4a_branch2a, use_global_stats=True,
                                            fix_gamma=False, eps = self.eps)
        scale4a_branch2a = bn4a_branch2a
        res4a_branch2a_relu = mx.symbol.Activation(name='res4a_branch2a_relu', data=scale4a_branch2a, act_type='relu')
        res4a_branch2b = mx.symbol.Convolution(name='res4a_branch2b', data=res4a_branch2a_relu, num_filter=256,
                                               pad=(1, 1), kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn4a_branch2b = mx.symbol.BatchNorm(name='bn4a_branch2b', data=res4a_branch2b, use_global_stats=True,
                                            fix_gamma=False, eps = self.eps)
        scale4a_branch2b = bn4a_branch2b
        res4a_branch2b_relu = mx.symbol.Activation(name='res4a_branch2b_relu', data=scale4a_branch2b, act_type='relu')

def get_detnet_backbone(self, data, is_train = True, with_dilated=True,  eps=1e-5):
        use_global_stats  = True#not is_train
        conv1 = mx.symbol.Convolution(name='conv1', data=data, num_filter=64, pad=(3, 3), kernel=(7, 7), stride=(2, 2), no_bias=True)
        bn_conv1 = mx.symbol.BatchNorm(name='bn_conv1', data=conv1, use_global_stats=use_global_stats, fix_gamma=False, eps=eps)
        scale_conv1 = bn_conv1
        conv1_relu = mx.symbol.Activation(name='conv1_relu', data=scale_conv1, act_type='relu')
        pool1 = mx.symbol.Pooling(name='pool1', data=conv1_relu, pad=(1, 1), kernel=(3, 3), stride=(2, 2), pool_type='max')
        res2a_branch1 = mx.symbol.Convolution(name='res2a_branch1', data=pool1, num_filter=256, pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn2a_branch1 = mx.symbol.BatchNorm(name='bn2a_branch1', data=res2a_branch1, use_global_stats=use_global_stats, fix_gamma=False, eps=eps)
        scale2a_branch1 = bn2a_branch1
        res2a_branch2a = mx.symbol.Convolution(name='res2a_branch2a', data=pool1, num_filter=64, pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn2a_branch2a = mx.symbol.BatchNorm(name='bn2a_branch2a', data=res2a_branch2a, use_global_stats=use_global_stats, fix_gamma=False, eps=eps)
        scale2a_branch2a = bn2a_branch2a
        res2a_branch2a_relu = mx.symbol.Activation(name='res2a_branch2a_relu', data=scale2a_branch2a, act_type='relu')
        res2a_branch2b = mx.symbol.Convolution(name='res2a_branch2b', data=res2a_branch2a_relu, num_filter=64, pad=(1, 1), kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn2a_branch2b = mx.symbol.BatchNorm(name='bn2a_branch2b', data=res2a_branch2b, use_global_stats=use_global_stats, fix_gamma=False, eps=eps)
        scale2a_branch2b = bn2a_branch2b
        res2a_branch2b_relu = mx.symbol.Activation(name='res2a_branch2b_relu', data=scale2a_branch2b, act_type='relu')
        res2a_branch2c = mx.symbol.Convolution(name='res2a_branch2c', data=res2a_branch2b_relu, num_filter=256,
                                               pad=(0, 0),
                                               kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn2a_branch2c = mx.symbol.BatchNorm(name='bn2a_branch2c', data=res2a_branch2c, use_global_stats=use_global_stats,
                                            fix_gamma=False, eps=eps)
        scale2a_branch2c = bn2a_branch2c

scale4b13_branch2a = bn4b13_branch2a
        res4b13_branch2a_relu = mx.symbol.Activation(name='res4b13_branch2a_relu', data=scale4b13_branch2a, act_type='relu')
        res4b13_branch2b = mx.symbol.Convolution(name='res4b13_branch2b', data=res4b13_branch2a_relu, num_filter=256,
                                                 pad=(1, 1), kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn4b13_branch2b = mx.symbol.BatchNorm(name='bn4b13_branch2b', data=res4b13_branch2b, use_global_stats=True,
                                              fix_gamma=False, eps=self.eps)
        scale4b13_branch2b = bn4b13_branch2b
        res4b13_branch2b_relu = mx.symbol.Activation(name='res4b13_branch2b_relu', data=scale4b13_branch2b, act_type='relu')
        res4b13_branch2c = mx.symbol.Convolution(name='res4b13_branch2c', data=res4b13_branch2b_relu, num_filter=1024,
                                                 pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b13_branch2c = mx.symbol.BatchNorm(name='bn4b13_branch2c', data=res4b13_branch2c, use_global_stats=True,
                                              fix_gamma=False, eps=self.eps)
        scale4b13_branch2c = bn4b13_branch2c
        res4b13 = mx.symbol.broadcast_add(name='res4b13', *[res4b12_relu, scale4b13_branch2c])
        res4b13_relu = mx.symbol.Activation(name='res4b13_relu', data=res4b13, act_type='relu')
        res4b14_branch2a = mx.symbol.Convolution(name='res4b14_branch2a', data=res4b13_relu, num_filter=256, pad=(0, 0),
                                                 kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b14_branch2a = mx.symbol.BatchNorm(name='bn4b14_branch2a', data=res4b14_branch2a, use_global_stats=True,
                                              fix_gamma=False, eps=self.eps)
        scale4b14_branch2a = bn4b14_branch2a
        res4b14_branch2a_relu = mx.symbol.Activation(name='res4b14_branch2a_relu', data=scale4b14_branch2a, act_type='relu')
        res4b14_branch2b = mx.symbol.Convolution(name='res4b14_branch2b', data=res4b14_branch2a_relu, num_filter=256,
                                                 pad=(1, 1), kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn4b14_branch2b = mx.symbol.BatchNorm(name='bn4b14_branch2b', data=res4b14_branch2b, use_global_stats=True,
                                              fix_gamma=False, eps=self.eps)
        scale4b14_branch2b = bn4b14_branch2b
        res4b14_branch2b_relu = mx.symbol.Activation(name='res4b14_branch2b_relu', data=scale4b14_branch2b, act_type='relu')
        res4b14_branch2c = mx.symbol.Convolution(name='res4b14_branch2c', data=res4b14_branch2b_relu, num_filter=1024,
                                                 pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b14_branch2c = mx.symbol.BatchNorm(name='bn4b14_branch2c', data=res4b14_branch2c, use_global_stats=True,
                                              fix_gamma=False, eps=self.eps)
        scale4b14_branch2c = bn4b14_branch2c

res3b3_branch2b_offset = mx.symbol.Convolution(name='res3b3_branch2b_offset', data=res3b3_branch2a_relu,
                                                           num_filter=72, pad=(1, 1), kernel=(3, 3), stride=(1, 1))
            res3b3_branch2b = mx.contrib.symbol.DeformableConvolution(name='res3b3_branch2b', data=res3b3_branch2a_relu,
                                                                      offset=res3b3_branch2b_offset,
                                                                      num_filter=128, pad=(1, 1), kernel=(3, 3),
                                                                      num_deformable_group=4,
                                                                      stride=(1, 1), no_bias=True)
        else:
            res3b3_branch2b = mx.symbol.Convolution(name='res3b3_branch2b', data=res3b3_branch2a_relu, num_filter=128,
                                                    pad=(1, 1), kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn3b3_branch2b = mx.symbol.BatchNorm(name='bn3b3_branch2b', data=res3b3_branch2b, use_global_stats=True,
                                             fix_gamma=False, eps=eps)
        scale3b3_branch2b = bn3b3_branch2b
        res3b3_branch2b_relu = mx.symbol.Activation(name='res3b3_branch2b_relu', data=scale3b3_branch2b,
                                                    act_type='relu')
        res3b3_branch2c = mx.symbol.Convolution(name='res3b3_branch2c', data=res3b3_branch2b_relu, num_filter=512,
                                                pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn3b3_branch2c = mx.symbol.BatchNorm(name='bn3b3_branch2c', data=res3b3_branch2c, use_global_stats=True,
                                             fix_gamma=False, eps=eps)
        scale3b3_branch2c = bn3b3_branch2c
        res3b3 = mx.symbol.broadcast_add(name='res3b3', *[res3b2_relu, scale3b3_branch2c])
        res3b3_relu = mx.symbol.Activation(name='res3b3_relu', data=res3b3, act_type='relu')
        res4a_branch1 = mx.symbol.Convolution(name='res4a_branch1', data=res3b3_relu, num_filter=1024, pad=(0, 0),
                                              kernel=(1, 1), stride=(2, 2), no_bias=True)
        bn4a_branch1 = mx.symbol.BatchNorm(name='bn4a_branch1', data=res4a_branch1, use_global_stats=True,
                                           fix_gamma=False, eps=eps)
        scale4a_branch1 = bn4a_branch1
        res4a_branch2a = mx.symbol.Convolution(name='res4a_branch2a', data=res3b3_relu, num_filter=256, pad=(0, 0),
                                               kernel=(1, 1), stride=(2, 2), no_bias=True)
        bn4a_branch2a = mx.symbol.BatchNorm(name='bn4a_branch2a', data=res4a_branch2a, use_global_stats=True,
                                            fix_gamma=False, eps=eps)
        scale4a_branch2a = bn4a_branch2a

res2a = mx.symbol.broadcast_add(name='res2a', *[scale2a_branch1, scale2a_branch2c])
        res2a_relu = mx.symbol.Activation(name='res2a_relu', data=res2a, act_type='relu')
        res2b_branch2a = mx.symbol.Convolution(name='res2b_branch2a', data=res2a_relu, num_filter=64, pad=(0, 0),
                                               kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn2b_branch2a = mx.symbol.BatchNorm(name='bn2b_branch2a', data=res2b_branch2a, use_global_stats=True,
                                            fix_gamma=False, eps=self.eps)
        scale2b_branch2a = bn2b_branch2a
        res2b_branch2a_relu = mx.symbol.Activation(name='res2b_branch2a_relu', data=scale2b_branch2a, act_type='relu')
        res2b_branch2b = mx.symbol.Convolution(name='res2b_branch2b', data=res2b_branch2a_relu, num_filter=64,
                                               pad=(1, 1),
                                               kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn2b_branch2b = mx.symbol.BatchNorm(name='bn2b_branch2b', data=res2b_branch2b, use_global_stats=True,
                                            fix_gamma=False, eps=self.eps)
        scale2b_branch2b = bn2b_branch2b
        res2b_branch2b_relu = mx.symbol.Activation(name='res2b_branch2b_relu', data=scale2b_branch2b, act_type='relu')
        res2b_branch2c = mx.symbol.Convolution(name='res2b_branch2c', data=res2b_branch2b_relu, num_filter=256,
                                               pad=(0, 0),
                                               kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn2b_branch2c = mx.symbol.BatchNorm(name='bn2b_branch2c', data=res2b_branch2c, use_global_stats=True,
                                            fix_gamma=False, eps=self.eps)
        scale2b_branch2c = bn2b_branch2c
        res2b = mx.symbol.broadcast_add(name='res2b', *[res2a_relu, scale2b_branch2c])
        res2b_relu = mx.symbol.Activation(name='res2b_relu', data=res2b, act_type='relu')
        res2c_branch2a = mx.symbol.Convolution(name='res2c_branch2a', data=res2b_relu, num_filter=64, pad=(0, 0),
                                               kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn2c_branch2a = mx.symbol.BatchNorm(name='bn2c_branch2a', data=res2c_branch2a, use_global_stats=True,
                                            fix_gamma=False, eps=self.eps)
        scale2c_branch2a = bn2c_branch2a
        res2c_branch2a_relu = mx.symbol.Activation(name='res2c_branch2a_relu', data=scale2c_branch2a, act_type='relu')
        res2c_branch2b = mx.symbol.Convolution(name='res2c_branch2b', data=res2c_branch2a_relu, num_filter=64,
                                               pad=(1, 1),
                                               kernel=(3, 3), stride=(1, 1), no_bias=True)

conv4_6_1x1_increase_bn = mx.symbol.BatchNorm(name='conv4_6_1x1_increase_bn', data=conv4_6_1x1_increase , use_global_stats=True, fix_gamma=False, eps=0.000100)
    conv4_6_1x1_increase_bn_scale = conv4_6_1x1_increase_bn
    conv4_6_global_pool = mx.symbol.Pooling(name='conv4_6_global_pool', data=conv4_6_1x1_increase_bn_scale , pooling_convention='full', global_pool=True, kernel=(1,1), pool_type='avg')
    conv4_6_1x1_down = mx.symbol.Convolution(name='conv4_6_1x1_down', data=conv4_6_global_pool , num_filter=64, pad=(0, 0), kernel=(1,1), stride=(1,1), no_bias=False)
    conv4_6_1x1_down_relu = mx.symbol.Activation(name='conv4_6_1x1_down_relu', data=conv4_6_1x1_down , act_type='relu')
    conv4_6_1x1_up = mx.symbol.Convolution(name='conv4_6_1x1_up', data=conv4_6_1x1_down_relu , num_filter=1024, pad=(0, 0), kernel=(1,1), stride=(1,1), no_bias=False)
    conv4_6_prob = mx.symbol.Activation(name='conv4_6_prob', data=conv4_6_1x1_up , act_type='sigmoid')
    if memonger:
        conv4_5_relu._set_attr(mirror_stage='True')
    conv4_6 = mx.sym.broadcast_mul(conv4_6_prob, conv4_6_1x1_increase_bn_scale) + conv4_5_relu
    conv4_6_relu = mx.symbol.Activation(name='conv4_6_relu', data=conv4_6 , act_type='relu')
    conv5_1_1x1_reduce = mx.symbol.Convolution(name='conv5_1_1x1_reduce', data=conv4_6_relu , num_filter=1024, pad=(0, 0), kernel=(1,1), stride=(1,1), no_bias=True)
    conv5_1_1x1_reduce_bn = mx.symbol.BatchNorm(name='conv5_1_1x1_reduce_bn', data=conv5_1_1x1_reduce , use_global_stats=True, fix_gamma=False, eps=0.000100)
    conv5_1_1x1_reduce_bn_scale = conv5_1_1x1_reduce_bn
    conv5_1_1x1_reduce_relu = mx.symbol.Activation(name='conv5_1_1x1_reduce_relu', data=conv5_1_1x1_reduce_bn_scale , act_type='relu')
    conv5_1_3x3 = mx.symbol.Convolution(name='conv5_1_3x3', data=conv5_1_1x1_reduce_relu , num_filter=1024, pad=(1, 1), kernel=(3,3), stride=(2,2), no_bias=True, num_group=32)
    conv5_1_3x3_bn = mx.symbol.BatchNorm(name='conv5_1_3x3_bn', data=conv5_1_3x3 , use_global_stats=True, fix_gamma=False, eps=0.000100)
    conv5_1_3x3_bn_scale = conv5_1_3x3_bn
    conv5_1_3x3_relu = mx.symbol.Activation(name='conv5_1_3x3_relu', data=conv5_1_3x3_bn_scale , act_type='relu')
    conv5_1_1x1_increase = mx.symbol.Convolution(name='conv5_1_1x1_increase', data=conv5_1_3x3_relu , num_filter=2048, pad=(0, 0), kernel=(1,1), stride=(1,1), no_bias=True)
    conv5_1_1x1_increase_bn = mx.symbol.BatchNorm(name='conv5_1_1x1_increase_bn', data=conv5_1_1x1_increase , use_global_stats=True, fix_gamma=False, eps=0.000100)
    conv5_1_1x1_increase_bn_scale = conv5_1_1x1_increase_bn
    conv5_1_global_pool = mx.symbol.Pooling(name='conv5_1_global_pool', data=conv5_1_1x1_increase_bn_scale , pooling_convention='full', global_pool=True, kernel=(1,1), pool_type='avg')
    conv5_1_1x1_down = mx.symbol.Convolution(name='conv5_1_1x1_down', data=conv5_1_global_pool , num_filter=128, pad=(0, 0), kernel=(1,1), stride=(1,1), no_bias=False)
    conv5_1_1x1_down_relu = mx.symbol.Activation(name='conv5_1_1x1_down_relu', data=conv5_1_1x1_down , act_type='relu')
    conv5_1_1x1_up = mx.symbol.Convolution(name='conv5_1_1x1_up', data=conv5_1_1x1_down_relu , num_filter=2048, pad=(0, 0), kernel=(1,1), stride=(1,1), no_bias=False)
    conv5_1_prob = mx.symbol.Activation(name='conv5_1_prob', data=conv5_1_1x1_up , act_type='sigmoid')
    conv5_1_1x1_proj = mx.symbol.Convolution(name='conv5_1_1x1_proj', data=conv4_6_relu , num_filter=2048, pad=(0, 0), kernel=(1,1), stride=(2,2), no_bias=True)
    conv5_1_1x1_proj_bn = mx.symbol.BatchNorm(name='conv5_1_1x1_proj_bn', data=conv5_1_1x1_proj , use_global_stats=True, fix_gamma=False, eps=0.000100)
    conv5_1_1x1_proj_bn_scale = conv5_1_1x1_proj_bn
    if memonger:
        conv5_1_1x1_proj_bn_scale._set_attr(mirror_stage='True')

scale3b3_branch2a = bn3b3_branch2a
        res3b3_branch2a_relu = mx.symbol.Activation(name='res3b3_branch2a_relu', data=scale3b3_branch2a, act_type='relu')
        res3b3_branch2b = mx.symbol.Convolution(name='res3b3_branch2b', data=res3b3_branch2a_relu, num_filter=128,
                                                pad=(1, 1), kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn3b3_branch2b = mx.symbol.BatchNorm(name='bn3b3_branch2b', data=res3b3_branch2b, use_global_stats=True,
                                             fix_gamma=False, eps=self.eps)
        scale3b3_branch2b = bn3b3_branch2b
        res3b3_branch2b_relu = mx.symbol.Activation(name='res3b3_branch2b_relu', data=scale3b3_branch2b, act_type='relu')
        res3b3_branch2c = mx.symbol.Convolution(name='res3b3_branch2c', data=res3b3_branch2b_relu, num_filter=512,
                                                pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn3b3_branch2c = mx.symbol.BatchNorm(name='bn3b3_branch2c', data=res3b3_branch2c, use_global_stats=True,
                                             fix_gamma=False, eps=self.eps)
        scale3b3_branch2c = bn3b3_branch2c
        res3b3 = mx.symbol.broadcast_add(name='res3b3', *[res3b2_relu, scale3b3_branch2c])
        res3b3_relu = mx.symbol.Activation(name='res3b3_relu', data=res3b3, act_type='relu')
        res4a_branch1 = mx.symbol.Convolution(name='res4a_branch1', data=res3b3_relu, num_filter=1024, pad=(0, 0),
                                              kernel=(1, 1), stride=(2, 2), no_bias=True)
        bn4a_branch1 = mx.symbol.BatchNorm(name='bn4a_branch1', data=res4a_branch1, use_global_stats=True, fix_gamma=False, eps=self.eps)
        scale4a_branch1 = bn4a_branch1
        res4a_branch2a = mx.symbol.Convolution(name='res4a_branch2a', data=res3b3_relu, num_filter=256, pad=(0, 0),
                                               kernel=(1, 1), stride=(2, 2), no_bias=True)
        bn4a_branch2a = mx.symbol.BatchNorm(name='bn4a_branch2a', data=res4a_branch2a, use_global_stats=True,
                                            fix_gamma=False, eps=self.eps)
        scale4a_branch2a = bn4a_branch2a
        res4a_branch2a_relu = mx.symbol.Activation(name='res4a_branch2a_relu', data=scale4a_branch2a, act_type='relu')
        res4a_branch2b = mx.symbol.Convolution(name='res4a_branch2b', data=res4a_branch2a_relu, num_filter=256, pad=(1, 1),
                                               kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn4a_branch2b = mx.symbol.BatchNorm(name='bn4a_branch2b', data=res4a_branch2b, use_global_stats=True,
                                            fix_gamma=False, eps=self.eps)
        scale4a_branch2b = bn4a_branch2b
        res4a_branch2b_relu = mx.symbol.Activation(name='res4a_branch2b_relu', data=scale4a_branch2b, act_type='relu')
        res4a_branch2c = mx.symbol.Convolution(name='res4a_branch2c', data=res4a_branch2b_relu, num_filter=1024, pad=(0, 0),

act_type='relu')
        res4b18_branch2c = mx.symbol.Convolution(name='res4b18_branch2c', data=res4b18_branch2b_relu, num_filter=1024,
                                                 pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b18_branch2c = mx.symbol.BatchNorm(name='bn4b18_branch2c', data=res4b18_branch2c, use_global_stats=True,
                                              fix_gamma=False, eps = self.eps)
        scale4b18_branch2c = bn4b18_branch2c
        res4b18 = mx.symbol.broadcast_add(name='res4b18', *[res4b17_relu, scale4b18_branch2c])
        res4b18_relu = mx.symbol.Activation(name='res4b18_relu', data=res4b18, act_type='relu')
        res4b19_branch2a = mx.symbol.Convolution(name='res4b19_branch2a', data=res4b18_relu, num_filter=256, pad=(0, 0),
                                                 kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b19_branch2a = mx.symbol.BatchNorm(name='bn4b19_branch2a', data=res4b19_branch2a, use_global_stats=True,
                                              fix_gamma=False, eps = self.eps)
        scale4b19_branch2a = bn4b19_branch2a
        res4b19_branch2a_relu = mx.symbol.Activation(name='res4b19_branch2a_relu', data=scale4b19_branch2a,
                                                     act_type='relu')
        res4b19_branch2b = mx.symbol.Convolution(name='res4b19_branch2b', data=res4b19_branch2a_relu, num_filter=256,
                                                 pad=(1, 1), kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn4b19_branch2b = mx.symbol.BatchNorm(name='bn4b19_branch2b', data=res4b19_branch2b, use_global_stats=True,
                                              fix_gamma=False, eps = self.eps)
        scale4b19_branch2b = bn4b19_branch2b
        res4b19_branch2b_relu = mx.symbol.Activation(name='res4b19_branch2b_relu', data=scale4b19_branch2b,
                                                     act_type='relu')
        res4b19_branch2c = mx.symbol.Convolution(name='res4b19_branch2c', data=res4b19_branch2b_relu, num_filter=1024,
                                                 pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b19_branch2c = mx.symbol.BatchNorm(name='bn4b19_branch2c', data=res4b19_branch2c, use_global_stats=True,
                                              fix_gamma=False, eps = self.eps)
        scale4b19_branch2c = bn4b19_branch2c
        res4b19 = mx.symbol.broadcast_add(name='res4b19', *[res4b18_relu, scale4b19_branch2c])
        res4b19_relu = mx.symbol.Activation(name='res4b19_relu', data=res4b19, act_type='relu')
        res4b20_branch2a = mx.symbol.Convolution(name='res4b20_branch2a', data=res4b19_relu, num_filter=256, pad=(0, 0),
                                                 kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b20_branch2a = mx.symbol.BatchNorm(name='bn4b20_branch2a', data=res4b20_branch2a, use_global_stats=True,

scale4b2_branch2b = bn4b2_branch2b
        res4b2_branch2b_relu = mx.symbol.Activation(name='res4b2_branch2b_relu', data=scale4b2_branch2b, act_type='relu')
        res4b2_branch2c = mx.symbol.Convolution(name='res4b2_branch2c', data=res4b2_branch2b_relu, num_filter=1024,
                                                pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b2_branch2c = mx.symbol.BatchNorm(name='bn4b2_branch2c', data=res4b2_branch2c, use_global_stats=True,
                                             fix_gamma=False, eps=self.eps)
        scale4b2_branch2c = bn4b2_branch2c
        res4b2 = mx.symbol.broadcast_add(name='res4b2', *[res4b1_relu, scale4b2_branch2c])
        res4b2_relu = mx.symbol.Activation(name='res4b2_relu', data=res4b2, act_type='relu')
        res4b3_branch2a = mx.symbol.Convolution(name='res4b3_branch2a', data=res4b2_relu, num_filter=256, pad=(0, 0),
                                                kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b3_branch2a = mx.symbol.BatchNorm(name='bn4b3_branch2a', data=res4b3_branch2a, use_global_stats=True,
                                             fix_gamma=False, eps=self.eps)
        scale4b3_branch2a = bn4b3_branch2a
        res4b3_branch2a_relu = mx.symbol.Activation(name='res4b3_branch2a_relu', data=scale4b3_branch2a, act_type='relu')
        res4b3_branch2b = mx.symbol.Convolution(name='res4b3_branch2b', data=res4b3_branch2a_relu, num_filter=256,
                                                pad=(1, 1), kernel=(3, 3), stride=(1, 1), no_bias=True)
        bn4b3_branch2b = mx.symbol.BatchNorm(name='bn4b3_branch2b', data=res4b3_branch2b, use_global_stats=True,
                                             fix_gamma=False, eps=self.eps)
        scale4b3_branch2b = bn4b3_branch2b
        res4b3_branch2b_relu = mx.symbol.Activation(name='res4b3_branch2b_relu', data=scale4b3_branch2b, act_type='relu')
        res4b3_branch2c = mx.symbol.Convolution(name='res4b3_branch2c', data=res4b3_branch2b_relu, num_filter=1024,
                                                pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b3_branch2c = mx.symbol.BatchNorm(name='bn4b3_branch2c', data=res4b3_branch2c, use_global_stats=True,
                                             fix_gamma=False, eps=self.eps)
        scale4b3_branch2c = bn4b3_branch2c
        res4b3 = mx.symbol.broadcast_add(name='res4b3', *[res4b2_relu, scale4b3_branch2c])
        res4b3_relu = mx.symbol.Activation(name='res4b3_relu', data=res4b3, act_type='relu')
        res4b4_branch2a = mx.symbol.Convolution(name='res4b4_branch2a', data=res4b3_relu, num_filter=256, pad=(0, 0),
                                                kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn4b4_branch2a = mx.symbol.BatchNorm(name='bn4b4_branch2a', data=res4b4_branch2a, use_global_stats=True,
                                             fix_gamma=False, eps=self.eps)

scale5b_branch2a = bn5b_branch2a
        res5b_branch2a_relu = mx.symbol.Activation(name='res5b_branch2a_relu', data=scale5b_branch2a, act_type='relu')
        res5b_branch2b = mx.symbol.Convolution(name='res5b_branch2b', data=res5b_branch2a_relu, num_filter=512,
                                               pad=(2, 2), kernel=(3, 3), dilate=(2, 2), stride=(1, 1), no_bias=True)
        bn5b_branch2b = mx.symbol.BatchNorm(name='bn5b_branch2b', data=res5b_branch2b, use_global_stats=True,
                                            fix_gamma=False, eps = self.eps)
        scale5b_branch2b = bn5b_branch2b
        res5b_branch2b_relu = mx.symbol.Activation(name='res5b_branch2b_relu', data=scale5b_branch2b, act_type='relu')
        res5b_branch2c = mx.symbol.Convolution(name='res5b_branch2c', data=res5b_branch2b_relu, num_filter=2048,
                                               pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn5b_branch2c = mx.symbol.BatchNorm(name='bn5b_branch2c', data=res5b_branch2c, use_global_stats=True,
                                            fix_gamma=False, eps = self.eps)
        scale5b_branch2c = bn5b_branch2c
        res5b = mx.symbol.broadcast_add(name='res5b', *[res5a_relu, scale5b_branch2c])
        res5b_relu = mx.symbol.Activation(name='res5b_relu', data=res5b, act_type='relu')
        res5c_branch2a = mx.symbol.Convolution(name='res5c_branch2a', data=res5b_relu, num_filter=512, pad=(0, 0),
                                               kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn5c_branch2a = mx.symbol.BatchNorm(name='bn5c_branch2a', data=res5c_branch2a, use_global_stats=True,
                                            fix_gamma=False, eps = self.eps)
        scale5c_branch2a = bn5c_branch2a
        res5c_branch2a_relu = mx.symbol.Activation(name='res5c_branch2a_relu', data=scale5c_branch2a, act_type='relu')
        res5c_branch2b = mx.symbol.Convolution(name='res5c_branch2b', data=res5c_branch2a_relu, num_filter=512,
                                               pad=(2, 2), kernel=(3, 3), dilate=(2, 2), stride=(1, 1), no_bias=True)
        bn5c_branch2b = mx.symbol.BatchNorm(name='bn5c_branch2b', data=res5c_branch2b, use_global_stats=True,
                                            fix_gamma=False, eps = self.eps)
        scale5c_branch2b = bn5c_branch2b
        res5c_branch2b_relu = mx.symbol.Activation(name='res5c_branch2b_relu', data=scale5c_branch2b, act_type='relu')
        res5c_branch2c = mx.symbol.Convolution(name='res5c_branch2c', data=res5c_branch2b_relu, num_filter=2048,
                                               pad=(0, 0), kernel=(1, 1), stride=(1, 1), no_bias=True)
        bn5c_branch2c = mx.symbol.BatchNorm(name='bn5c_branch2c', data=res5c_branch2c, use_global_stats=True,
                                            fix_gamma=False, eps = self.eps)
        scale5c_branch2c = bn5c_branch2c