model.txt

EfficientDet(
  (stages): HybridSequential(
    (0): HybridSequential(
      (0): Conv2D(3 -> 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=32)
      (2): Activation(
        (act): Swish(
        
        )
      )
      (3): HybridSequential(
        (0): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(1 -> 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=32)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(32 -> 8, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(8 -> 32, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (4): Conv2D(32 -> 16, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (5): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=16)
          )
        )
      )
      (4): HybridSequential(
        (0): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(16 -> 96, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=96)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 96, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=96, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=96)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(96 -> 24, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(24 -> 96, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(96 -> 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=24)
          )
        )
        (1): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(24 -> 144, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 144, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=144, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(144 -> 40, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(40 -> 144, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(144 -> 24, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=24)
          )
        )
      )
      (5): HybridSequential(
        (0): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(24 -> 144, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 144, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=144, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(144 -> 40, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(40 -> 144, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(144 -> 40, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=40)
          )
        )
        (1): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(40 -> 240, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=240)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=240)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(240 -> 64, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(64 -> 240, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(240 -> 40, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=40)
          )
        )
      )
    )
    (1): HybridSequential(
      (0): HybridSequential(
        (0): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(40 -> 240, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=240)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 240, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=240, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=240)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(240 -> 64, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(64 -> 240, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(240 -> 80, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=80)
          )
        )
        (1): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(80 -> 480, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=480)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 480, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=480, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=480)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(480 -> 120, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(120 -> 480, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(480 -> 80, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=80)
          )
        )
        (2): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(80 -> 480, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=480)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 480, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=480, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=480)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(480 -> 120, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(120 -> 480, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(480 -> 80, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=80)
          )
        )
      )
      (1): HybridSequential(
        (0): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(80 -> 480, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=480)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 480, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=480, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=480)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(480 -> 120, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(120 -> 480, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(480 -> 112, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=112)
          )
        )
        (1): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(112 -> 672, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=672)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 672, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=672, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=672)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(672 -> 168, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(168 -> 672, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(672 -> 112, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=112)
          )
        )
        (2): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(112 -> 672, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=672)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 672, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=672, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=672)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(672 -> 168, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(168 -> 672, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(672 -> 112, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=112)
          )
        )
      )
    )
    (2): HybridSequential(
      (0): HybridSequential(
        (0): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(112 -> 672, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=672)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 672, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=672, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=672)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(672 -> 168, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(168 -> 672, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(672 -> 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=192)
          )
        )
        (1): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(192 -> 1152, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 1152, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=1152, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(1152 -> 288, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(288 -> 1152, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(1152 -> 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=192)
          )
        )
        (2): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(192 -> 1152, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 1152, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=1152, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(1152 -> 288, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(288 -> 1152, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(1152 -> 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=192)
          )
        )
        (3): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(192 -> 1152, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 1152, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=1152, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(1152 -> 288, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(288 -> 1152, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(1152 -> 192, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=192)
          )
        )
      )
      (1): HybridSequential(
        (0): LinearBottleneck(
          (dropout): Dropout(p = 0.2, axes=(1, 2, 3))
          (out): HybridSequential(
            (0): Conv2D(192 -> 1152, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
            (2): Activation(
              (act): Swish(
              
              )
            )
            (3): Conv2D(1 -> 1152, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1152, bias=False)
            (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
            (5): Activation(
              (act): Swish(
              
              )
            )
            (6): _SE(
              (pool): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
              (conv1): Conv2D(1152 -> 288, kernel_size=(1, 1), stride=(1, 1))
              (act1): Activation(
                (act): Activation(relu)
              )
              (conv2): Conv2D(288 -> 1152, kernel_size=(1, 1), stride=(1, 1))
              (act2): Activation(
                (act): HardSigmoid(
                  (act): ReLU6(
                  
                  )
                )
              )
            )
            (7): Conv2D(1152 -> 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
            (8): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=320)
          )
        )
      )
    )
  )
  (proj_convs): HybridSequential(
    (0): HybridSequential(
      (0): Conv2D(40 -> 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
      (2): Activation(
        (act): Swish(
        
        )
      )
    )
    (1): HybridSequential(
      (0): Conv2D(112 -> 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
      (2): Activation(
        (act): Swish(
        
        )
      )
    )
    (2): HybridSequential(
      (0): Conv2D(320 -> 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
      (2): Activation(
        (act): Swish(
        
        )
      )
    )
    (3): HybridSequential(
      (0): Conv2D(320 -> 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
      (2): Activation(
        (act): Swish(
        
        )
      )
    )
    (4): HybridSequential(
      (0): Conv2D(320 -> 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
      (2): Activation(
        (act): Swish(
        
        )
      )
    )
  )
  (fpns): HybridSequential(
    (0): BiFPN(
      (convs): HybridSequential(
        (0): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (1): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (2): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (3): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (4): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (5): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (6): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (7): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
      )
    )
    (1): BiFPN(
      (convs): HybridSequential(
        (0): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (1): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (2): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (3): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (4): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (5): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (6): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (7): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
      )
    )
    (2): BiFPN(
      (convs): HybridSequential(
        (0): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (1): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (2): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (3): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (4): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (5): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (6): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
        (7): SeparableConvBlock(
          (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
          (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
          (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
          (act): Activation(
            (act): Swish(
            
            )
          )
        )
      )
    )
  )
  (anchor_generators): HybridSequential(
    (0): AnchorGenerator(
    
    )
    (1): AnchorGenerator(
    
    )
    (2): AnchorGenerator(
    
    )
    (3): AnchorGenerator(
    
    )
    (4): AnchorGenerator(
    
    )
  )
  (cls_net): OutputSubnet(
    (body): HybridSequential(
      (0): SeparableConvBlock(
        (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
        (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
        (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
        (act): Activation(
          (act): Swish(
          
          )
        )
      )
      (1): SeparableConvBlock(
        (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
        (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
        (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
        (act): Activation(
          (act): Swish(
          
          )
        )
      )
      (2): SeparableConvBlock(
        (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
        (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
        (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
        (act): Activation(
          (act): Swish(
          
          )
        )
      )
    )
    (output): SeparableConvBlock(
      (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
      (pointwise_conv): Conv2D(64 -> 45, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (box_net): OutputSubnet(
    (body): HybridSequential(
      (0): SeparableConvBlock(
        (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
        (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
        (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
        (act): Activation(
          (act): Swish(
          
          )
        )
      )
      (1): SeparableConvBlock(
        (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
        (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
        (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
        (act): Activation(
          (act): Swish(
          
          )
        )
      )
      (2): SeparableConvBlock(
        (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
        (pointwise_conv): Conv2D(64 -> 64, kernel_size=(1, 1), stride=(1, 1))
        (norm): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
        (act): Activation(
          (act): Swish(
          
          )
        )
      )
    )
    (output): SeparableConvBlock(
      (depthwise_conv): Conv2D(1 -> 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64, bias=False)
      (pointwise_conv): Conv2D(64 -> 36, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (bbox_decoder): NormalizedBoxCenterDecoder(
  
  )
  (cls_decoder): MultiPerClassDecoder(
  
  )
)
if __name__ == '__main__':
    args = parse_args()
    # context list
    ctx = [mx.gpu(int(i)) for i in args.gpus.split(',') if i.strip()]
    ctx = [mx.cpu()] if not ctx else ctx
    print("ctx=", ctx)
    
    # grab some image if not specified
    if not args.images.strip():
        gcv.utils.download("https://cloud.githubusercontent.com/assets/3307514/" +
            "20012568/cbc2d6f6-a27d-11e6-94c3-d35a9cb47609.jpg", 'street.jpg')
        image_list = ['street.jpg']
    else:
        image_list = [x.strip() for x in args.images.split(',') if x.strip()]

    if args.dataset.lower() == 'coco':
        from gluoncv.data import COCODetection
        classes = COCODetection.CLASSES
    elif args.dataset.lower() == 'voc':
        from gluoncv.data import VOCDetection
        classes = VOCDetection.CLASSES
    elif args.dataset.lower() == 'nzrc':
        classes = ['No Damage', 'Minor Damage', 'Major Damage', 'Destroyed']
    else:
        raise NotImplementedError('Dataset: {} not implemented.'.format(args.dataset))

    net = get_efficientdet(args.network, classes, pretrained_base=False)
    net.load_parameters(args.pretrained_path)
    net.set_nms(0.45, 200)
    net.collect_params().reset_ctx(ctx = ctx)

if __name__ == '__main__':