Closed czh1024 closed 2 years ago
Thanks!
@.***
From: Gabriele Cesa Date: 2022-02-05 01:17 To: QUVA-Lab/e2cnn_experiments CC: zihaochen; Mention Subject: Re: [QUVA-Lab/e2cnn_experiments] how to input 2 channels img? (Issue #1) Hi @czh1024 I am sorry, I did not see this issue before :( I think this is what you are asking: in_type = FieldType(gs, [gs.trivial_repr]*c)
where gs is the gspace you are interested in and c is the number of input channels (e.g. 2 in your case). You can then use in_type to define your first convolution layer as in most models in this repository. Does this help? Gabriele — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>
Hi @czh1024
I am sorry, I did not see this issue before :(
I think this is what you are asking:
where
gs
is the gspace you are interested in andc
is the number of input channels (e.g. 2 in your case). You can then usein_type
to define your first convolution layer as in most models in this repository.Does this help?
Gabriele