-
Notifications
You must be signed in to change notification settings - Fork 5
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
updated fMRI denoiser model and moved fMRI example to other (as it is…
… not in the same state as the ephys and ophys)
- Loading branch information
Showing
7 changed files
with
51 additions
and
33 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,65 +1,83 @@ | ||
| function net = fmriDenoiserNetwork(modelPath, pretrainedNetwork) | ||
| %fmriDenoiserNetwork - generate MATLAB Version of the DeepInterpolation fMRI-Denoiser | ||
| function net = fmriDenoiserNetwork(preN) | ||
| %FMRIDENOISERNETWORK - MATLAB Version of the DeepInterpolation fMRI-Denoiser | ||
|
|
||
| lgraph = layerGraph(); | ||
| resize = true; | ||
|
|
||
| tempLayers = [ | ||
| image3dInputLayer([7 7 7 5],"Name","image3dinput",'Normalization','none'); | ||
| convolution3dLayer([3 3 3],8,"Name","conv3d","Padding","same","Weights",mywini(1),"Bias",mybini(1)); | ||
| reluLayer("Name","relu1")]; | ||
| lgraph = addLayers(lgraph,tempLayers); | ||
| lgraph = layerGraph(); | ||
| poolsize = [2 2 2]; | ||
| poolargs={"Stride",poolsize}; | ||
| convargs={"Padding","same"}; | ||
| resizeLayers1 = [ | ||
| resize3dLayer("Scale",2,"Name","up1"); | ||
| functionLayer(@zeropad,"Name","pad1",Formattable=true,Acceleratable=true)]; | ||
| resizeLayers2 = [ | ||
| resize3dLayer("Scale",2,"Name","up2") | ||
| functionLayer(@zeropad,"Name","pad2",Formattable=true,Acceleratable=true)]; | ||
| cc1="pad1"; cc2="pad2"; | ||
|
|
||
| tempLayers = [ | ||
| maxPooling3dLayer([3 3 3],"Name","pool1","Padding","same") | ||
| convolution3dLayer([3 3 3],16,"Name","conv3d_1","Padding","same","Weights",mywini(2),"Bias",mybini(2)) | ||
| reluLayer("Name","relu2")]; | ||
| lgraph = addLayers(lgraph,tempLayers); | ||
| if ~resize | ||
| poolargs={"Padding","same"}; | ||
| resizeLayers1=[]; resizeLayers2=[]; | ||
| cc1="relu3"; cc2="relu4"; | ||
| end | ||
|
|
||
| tempLayers = [ | ||
| maxPooling3dLayer([3 3 3],"Name","pool2","Padding","same") | ||
| convolution3dLayer([3 3 3],32,"Name","conv3d_2","Padding","same","Weights",mywini(3),"Bias",mybini(3)) | ||
| reluLayer("Name","relu3")]; | ||
| image3dInputLayer([7 7 7 5],"Name","image3dinput","Normalization","none") | ||
| convolution3dLayer([3 3 3],8,"Name","conv1",convargs{:},"Weights",mywini(1,preN),"Bias",mybini(1,preN)) | ||
| reluLayer("Name","relu1") | ||
| maxPooling3dLayer(poolsize,"Name","pool1",poolargs{:}) | ||
| convolution3dLayer([3 3 3],16,"Name","conv2",convargs{:},"Weights",mywini(2,preN),"Bias",mybini(2,preN)) | ||
| reluLayer("Name","relu2") | ||
| maxPooling3dLayer(poolsize,"Name","pool2",poolargs{:}) | ||
| convolution3dLayer([3 3 3],32,"Name","conv3",convargs{:},"Weights",mywini(3,preN),"Bias",mybini(3,preN)) | ||
| reluLayer("Name","relu3") | ||
| resizeLayers1]; | ||
| lgraph = addLayers(lgraph,tempLayers); | ||
|
|
||
| tempLayers = [ | ||
| depthConcatenationLayer(2,"Name","conc_up_1") | ||
| convolution3dLayer([3 3 3],16,"Name","conv3d_3","Padding","same","Weights",mywini(4),"Bias",mybini(4)) | ||
| reluLayer("Name","relu4")]; | ||
| convolution3dLayer([3 3 3],16,"Name","conv4",convargs{:},"Weights",mywini(4,preN),"Bias",mybini(4,preN)) | ||
| reluLayer("Name","relu4") | ||
| resizeLayers2]; | ||
| lgraph = addLayers(lgraph,tempLayers); | ||
| lgraph = connectLayers(lgraph,cc1,"conc_up_1/in1"); | ||
| lgraph = connectLayers(lgraph,"relu2","conc_up_1/in2"); | ||
|
|
||
| tempLayers = [ | ||
| depthConcatenationLayer(2,"Name","conc_up_2") | ||
| convolution3dLayer([3 3 3],8,"Name","conv3d_4","Padding","same","Weights",mywini(5),"Bias",mybini(5)) | ||
| convolution3dLayer([3 3 3],8,"Name","conv5",convargs{:},"Weights",mywini(5,preN),"Bias",mybini(5,preN)) | ||
| reluLayer("Name","relu5") | ||
| convolution3dLayer([1 1 1],1,"Name","conv3d_5","Padding","same","Weights",mywini(6),"Bias",mybini(6)) | ||
| convolution3dLayer([1 1 1],1,"Name","out",convargs{:},"Weights",mywini(6,preN),"Bias",mybini(6,preN)) | ||
| regressionLayer("Name","out_r")]; | ||
| lgraph = addLayers(lgraph,tempLayers); | ||
| lgraph = connectLayers(lgraph,cc2,"conc_up_2/in1"); | ||
| lgraph = connectLayers(lgraph,"relu1","conc_up_2/in2"); | ||
|
|
||
| clear tempLayers; | ||
|
|
||
| lgraph = connectLayers(lgraph,"relu1","pool1"); | ||
| lgraph = connectLayers(lgraph,"relu1","conc_up_2/in2"); | ||
| lgraph = connectLayers(lgraph,"relu2","pool2"); | ||
| lgraph = connectLayers(lgraph,"relu2","conc_up_1/in2"); | ||
| lgraph = connectLayers(lgraph,"relu3","conc_up_1/in1"); | ||
| lgraph = connectLayers(lgraph,"relu4","conc_up_2/in1"); | ||
|
|
||
| net = assembleNetwork(lgraph); | ||
| save(modelPath,"net") | ||
| end | ||
|
|
||
|
|
||
| %% Local Functions | ||
|
|
||
| function w = mywini(ilayer) | ||
| function w = mywini(ilayer, preN) | ||
| lwlnames = {'conv3d','conv3d_1','conv3d_2','conv3d_3','conv3d_4','conv3d_5'}; %layers_with_learnables | ||
| lname = lwlnames{ilayer}; | ||
| thisweights = h5read(pretrainedNetwork,strcat('/model_weights/',lname,'/',lname,'/kernel:0')); | ||
| thisweights = h5read(preN,strcat('/model_weights/',lname,'/',lname,'/kernel:0')); | ||
| w = permute(thisweights,[5,4,3,2,1]); | ||
| end | ||
|
|
||
| function b = mybini(ilayer) | ||
| function b = mybini(ilayer, preN) | ||
| lwlnames = {'conv3d','conv3d_1','conv3d_2','conv3d_3','conv3d_4','conv3d_5'}; %layers_with_learnables | ||
| lwldims = [8 16 32 16 8 1]; | ||
| lname = lwlnames{ilayer}; | ||
| b = h5read(pretrainedNetwork,strcat('/model_weights/',lname,'/',lname,'/bias:0')); | ||
| b = h5read(preN,strcat('/model_weights/',lname,'/',lname,'/bias:0')); | ||
| b = reshape(b,[1 1 1 lwldims(ilayer)]); | ||
| end | ||
|
|
||
| function out = zeropad(in) | ||
| m = extractdata(in); | ||
| n = padarray(m,[1 1 1],0,"post"); | ||
| out = dlarray(n); | ||
| end |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file added
BIN
+408 KB
...dModels/2020_08_28_00_25_fmri_unet_denoiser_mean_absolute_error_2020_08_28_00_25_model.h5
Binary file not shown.
Binary file not shown.