Train deeplabv3 on your own dataset

[Vishrut10 (Vishrut)]

Thanks a lot for your reply.
Now I am stuck on the following:

Question about output and label channels in semantic segmentation

I use some nets,FCN8 ,SegNet for semantic segmentation .The trouble follow: all of the nets I used,The last layers of this net output the feature maps is (1,22,256,256),why not (1,3,256,256)? and another question is that the labels size is (1,1,256,256),why not(1,3,256,256)? please help me, I a new gay

Essentially converting the [batch_size, nb_classes, height, width] from output with 12 classes.

predictions[‘out’]
Out[76]:
tensor([[[[-0.0465, -0.0465, -0.0465, …, -0.0556, -0.0556, -0.0556],
[-0.0465, -0.0465, -0.0465, …, -0.0556, -0.0556, -0.0556],
[-0.0465, -0.0465, -0.0465, …, -0.0556, -0.0556, -0.0556],
…,

predictions[‘out’].size()
Out[77]: torch.Size([1, 12, 300, 400])