Conscious Artificial Intelligence Keras to CAI

[Dzandaa]

Hi everybody.

Using Lazarus 2.0.12 in Windows and Linux.

I try to translate a little keras python  test to CAI:

Code: Python  [Select][+][-]

1. import cv2
2. import keras
3. from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense
4. import numpy as np
5. import  matplotlib.pyplot as plt
6. import tensorflow as tf
7.  
8. img = cv2.imread("Funny_Ben.jpg")
9. img = cv2.resize(img, (224, 224))
10.  
11. model = keras.Sequential()
12.  
13. # Input
14. # 224x224x3
15.  
16. # Block 1
17. model.add(Conv2D(64, kernel_size=(3, 3), padding= "same", activation="relu", input_shape=(224, 224, 3)))
18. # 224x224x64
19. model.add(Conv2D(64, kernel_size=(3, 3), padding= "same", activation="relu"))
20. # 224x224x64
21. model.add(MaxPooling2D((2, 2), strides=(2,2)))
22. # 112x112x64
23.  
24. model.build()
25. model.summary()
26.  
27. # Result
28.  
29. result = model.predict(np.array([img]))
30.  
31. # Feature list
32. for i in range(64):
33.     feature_img = result[0, :, :, i]
34.     ax: object = plt.subplot(8, 8, i+1)
35.     ax.set_xticks([])
36.     ax.set_yticks([])
37.     plt.imshow(feature_img, cmap="gray")
38. plt.show()
39.  

what I have so far:

Code: Pascal  [Select][+][-]

1. OpenPicture: TOpenPictureDialog;  
2. InputVolume, PredictedVolume: TNNetVolume;
3. ImgInput: TImage;
4. NN: TNNet;  
5.  
6.  
7. procedure TestForm.BTestClick(Sender: TObject);
8. var
9.    Res: Boolean;
10.    TmpVolume: TNNetVolume;
11. begin
12.   InputVolume := TNNetVolume.Create();
13.   PredictedVolume := TNNetVolume.Create(1);
14.  
15.   if(OpenPicture.Execute) then
16.   begin
17.     TmpVolume := TNNetVolume.Create();
18.     Res := LoadImageFromFileIntoVolume(OpenPicture.FileName, TmpVolume);
19.     if(Res) then
20.     begin
21.      ImgInput.Width := 224;
22.      ImgInput.Height := 224;
23.      InputVolume.CopyResizing(TmpVolume, 224, 224);
24.      LoadVolumeIntoTImage(InputVolume, ImgInput);   // OK
25.  
26.      NN := TNNet.Create();
27.    
28.      NN.AddLayer(TNNetInput.Create(224, 224, 3)); // 224x224x3 Input Image
29.    
30.      // Block 1
31.    
32.      // 224x224x3
33.    
34.      NN.AddLayer(TNNetConvolutionReLU.Create({Features=}64, {FeatureSize=}3, {Padding=}0, {Stride=}0));
35.      // 224x224x64
36.      NN.AddLayer(TNNetConvolutionReLU.Create({Features=}64, {FeatureSize=}3, {Padding=}1, {Stride=}1));
37.      // 224x224x64
38.      NN.AddLayer(TNNetMaxPool.Create(2));                                                                      
39.      // 112x112x64
40.      
41.      NN.SetLearningRate(0.01,0.8);
42.      
43.      NN.Compute(InputVolume);
44.      NN.GetOutput(PredictedVolume);
45.      
46.      // Here I want to display the Channels list from PredictedVolume in a TImage, but I don't know how...    
47.            
48.              
49.     end;
50.    
51.     TmpVolume.Free;
52.   end;
53.   InputVolume.Free;
54.   PredictedVolume.Free;
55.   NN.Free;
56. end;                                  
57.  

I don't know if it is the right way to do that...

Any Help?

Thank you.

[schuler]

Hello @Dzandaa,
I presume that you intend to plot each channel of the activation map (not the features - neuronal weights).

To everyone else reading this post, if you do not know what an activation map is, you can have a look at https://towardsdatascience.com/activation-maps-for-deep-learning-models-in-a-few-lines-of-code-ed9ced1e8d21 .

@Dzandaa, on your Python code, you are extracting a channel of the activation map: feature_img = result[0, :, :, i] .

As this has the "channel last" notation, i is the channel.

This is what I would do in Pascal:

To produce an image, you need 3 channels (RGB). So, I would first produce a volume with 3 channels by copying the same channel with neuralvolume TVolume.CopyChannels(Original: TVolume; aChannels: array of integer) . In your code, it would look like this:

Code: Pascal  [Select][+][-]

1. ImageVolume.CopyChannels(PredictedVolume, [i,i,i]);

As you have a ReLU, you won't have negative values for us to care about.

Then, you'll need to scale the volume to [0, 255]:

Code: Pascal  [Select][+][-]

1. ImageVolume.NormalizeMax(255);

Then, to transform the ImageVolume into an image, you can call either SaveImageFromVolumeIntoFile or LoadVolumeIntoImage from neuraldatasets unit.

Alternatively, if your form has a TImage, you can call LoadVolumeIntoTImage from the unit neuralvolumev instead of calling SaveImageFromVolumeIntoFile or LoadVolumeIntoImage .

If, instead of plotting channels, you really intend to plot features, you can look at https://sourceforge.net/p/cai/svncode/HEAD/tree/trunk/lazarus/experiments/visualCifar10BatchUpdate/ . In this source code example, the features of the first layer are plotted.

Does this reply solve the question?

[Dzandaa]

Hi Joao Paulos,

Thank you very much.

This is just what I want, I will try it tomorrow.

[Dzandaa]

Hi,

It's working!!!

Thank you very much.

next Step OpenCL

B->