AdaptiveAvgPool2d

[shrutishrestha]

Why is this nn.AdaptiveAvgPool2d(1) done here?

class ASPPPooling(nn.Sequential):

def __init__(self, in_channels, out_channels):
    super(ASPPPooling, self).__init__(
        nn.AdaptiveAvgPool2d(1),
        nn.Conv2d(in_channels, out_channels, 1, bias=False),
        nn.BatchNorm2d(out_channels),
        nn.ReLU(inplace=True))

def forward(self, x):
    size = x.shape[-2:]
    x = super(ASPPPooling, self).forward(x)
    return F.interpolate(x, size=size, mode='bilinear', align_corners=False)

I am doing segmentation task and this abive pooling changes my output from torch.Size([1, 256, 16, 16]) to torch.Size([1, 256, 1, 1])
giving the error,
"Expected more than 1 value per channel when training, got input size torch.Size([1, 256, 1, 1])"

What could have gone wrong?

[Haienzi]

I have the same problem. How did you solve it？

[Nuller-CV]

I have the same problem. How did you solve it？

[shrutishrestha]

I have the same problem. How did you solve it？

Hi, how do you want your output? The line "nn.AdaptiveAvgPool2d(1)" means 1 is the output size and your output will be of height 1 and width 1.

[VainF]

BatchNorm requires at least 2 values per channel for mean and variance estimation. So, in training mode, please make sure that your batch_size is larger than 1.

[DISAPPEARED13]

what if batch_size is 1, can I just replace nn.BatchNorm with nn.InstanceNorm?