search

xingyifei2016 / RotLieNet

15 stars 4 forks source link

What are the components of each npy file? #2

Open GanggangDong opened 4 years ago

GanggangDong commented 4 years ago

Hi, Dr Xing, there are 5 sets of data in each npy file. I find the last axis (data[4, :, :]) is the magnitude. What are the data of other 4 axis? Thank you very much.

xingyifei2016 commented 4 years ago

Hello Ganggang, the input to the network is of tensor of the form [Batch, 2, 1, spatial x, spatial y] where the 2 corresponds to the polar coordinates of each complex value.

GanggangDong commented 4 years ago

Thank you very much. The polar coordinates of each complex value are magnitude and phase, is it ?

zakizuberiaria commented 4 years ago

@xingyifei2016 In addition the the question above, it appears that the first channel is the arccos() of the raw data file. Could you please explain why this is done?

Potentially is the calculation of the phase but that would assume that the first channel raw data is the real component divided by the magnitude.

mtlouie-unm commented 4 years ago

@GanggangDong @zakizuberiaria Were you able to figure out what data is being represented at each axis?

zakizuberiaria commented 4 years ago

I was able to get it running using the form suggested by @xingyifei2016 [Batch, 2, 1, spatial x, spatial y] where the 2nd dimension is mag and phase (in that order) as @GanggangDong suggested.

mtlouie-unm commented 4 years ago

So to clarify. In the 2nd dimension, we first have the magnitude, which is given as [4, :, :] the npy files. But for the phase we are computing the arccos, right? In the npy files, which index on the first dimension is the phase data?

GanggangDong commented 4 years ago

@mtlouie-unm @zakizuberiaria I am sorry. I really do not know what the remaining components denote. I am sure that the last channel is the magnitude.

GanggangDong commented 4 years ago

It is obvious that, component1=component4=arcsin(angle), specifically, component5.*sin(component1) = real component of SAR data the first component are as same as the fourth component, while the second component are different from the third component.

mtlouie-unm commented 4 years ago

@GanggangDong Ah I see it now. Thanks for the clarification!

18120803859 commented 5 months ago

Hello Ganggang, the input to the network is of tensor of the form [Batch, 2, 1, spatial x, spatial y] where the 2 corresponds to the polar coordinates of each complex value.

Dear Dr Xing，for the MSTAR experiment, the input of the network is of tensor of the form [Batch, 2, 1, spatial x, spatial y]. I would like to know for RadioML what the size of the input of the network? Is [Batch, 2, 1, 1, 128]?

xingyifei2016 commented 5 months ago

Hello,I believe it is indeed single channel and 128 dimension.On Jan 24, 2024, at 11:25 AM, 18120803859 @.***> wrote:

Hello Ganggang, the input to the network is of tensor of the form [Batch, 2, 1, spatial x, spatial y] where the 2 corresponds to the polar coordinates of each complex value.

Dear Dr Xing，for the MSTAR experiment, the input of the network is of tensor of the form [Batch, 2, 1, spatial x, spatial y]. I would like to know for RadioML what the size of the input of the network? Is [Batch, 2, 1, 1, 128]?

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: @.***>