search

zhaomin0101 / 3DCNN-HU

Implementation of paper "Hyperspectral Unmixing for Additive Nonlinear Models With a 3-D-CNN Autoencoder Network" (T-GRS 2022)
https://ieeexplore.ieee.org/document/9503107
MIT License
12 stars 1 forks source link

What parameters need to be changed and how. #1

Closed LongyuanCai closed 1 year ago

LongyuanCai commented 1 year ago

Hello I am a novice in this field. I've been trying to debug your program recently. If I have my own dataset, can you please tell me what parameters and code I need to change? Thank you.

LongyuanCai commented 1 year ago

Hello The specific problem is that samson's data is "data_samson.mat" (9025,156), but the data input into the model framework is "data_samson_patch_3.mat"(9025,9,156). How should I operate to obtain the latter data? Your paper does not mention how to operate. I'd love to hear from you. Thank you.

zhaomin0101 commented 1 year ago

Hello The specific problem is that samson's data is "data_samson.mat" (9025,156), but the data input into the model framework is "data_samson_patch_3.mat"(9025,9,156). How should I operate to obtain the latter data? Your paper does not mention how to operate. I'd love to hear from you. Thank you.

clear all clc data = importdata('data_linear.mat'); [a,b]=size(data); h = 100; w = 100; data = hyperConvert3d(data, h, w, a);

% window_size_3 data_new = [data(:,1,:),data,data(:,w,:)]; A =[data(1,1,:), data(1,:,:),data(h,1,:)]; B =[data(1,h,:), data(h,:,:),data(h,w,:)]; data_new = [A;data_new;B]; data_kuai = zeros(h*w,9,a); k = 1; for i=2:h+1 for j=2:w+1 C = data_new(i-1:i+1,j-1:j+1,:); C = reshape(C,[9,a]); %C = patch_sad(C,9); C = reshape(C,[1,9,a]); data_kuai(k,:,:)=C; k = k+1; end end data_patch =single(data_kuai);

LongyuanCai commented 1 year ago

Thank you very much for your answer. There is another question whether the hyperConvert3d function has the form of the following code.

function [img] = hyperConvert3d(img, h, w, numBands) if (ndims(img) ~= 2) error('Input image must be p x N.'); end [numBands, N] = size(img); if (1 == N) img = reshape( img, h, w); else img = reshape( img.', h, w, numBands); end return;