One can directly read GPR data and generate the images.
if one can dirctly get the plot image and save it, you can reduce the cpu excute time.
It costs my PC about 2 minutes to generate 200 images.
%generate b-scan
filename='G:\void_data\IRS\Dandong3\'; % dictory
str='Paodao3-'
for i=1:2
scale=rand*0.8+0.1;%plot scale
filename1=strcat(filename,'\',str,num2str(i),'-',num2str(ceil(scale*10)),'.jpg');
% figure
GPRBscanPlot1( TraceData,Sample_Fs,Sample_Space, startnum,scale );
set(0,'defaultfigurecolor','w');%background set to white
% saveas(1,filename1);%, 'jpg');
% I=imread(filename1);%,'.jpg'));% read the saved file
% image(I)
F=getframe(gcf); %convert plot to image
I=frame2im(F);
% image(I)
I=I(1:270,110:1000,:);%cut the rest white unuse area
I=imresize(I,[300,1090]);%resize it
imwrite(I,filename1); %write it again to resize to [300,1090]in height and width
% end
%
end
Here is the result
Then resize them to feed into your DL model
The test results