1 Convolutional Neural Network of «Supervision group» 2

1 Convolutional Neural Network of «Supervision group»

2

3 Feedforward convolutional network of “Supervision group”

4 Input Feedforward neural networks

5 Convolutional filtering Input Feedforward Feeedforward

6 Convolutional network of Kunihiko Fukushima Convolutional filtering Image structural synthesis Examples of image matching

7 Schematic structure of eye In such way the physiologists test the outputs of neural cells probing rod of oscillograph dendrites axon core mem-brane synapse Neuron cells and interneuronic interactions pigment cells rod cells conus cells horizontal cells amacrinal cells bipolar cells ganglionic cells Neural layers of retina cornea retina vitreous body crystalline lens

8 Simple cells Complex cells Complex cells

9

10

11 Weight Y X X Positions of the extremes of local difference Information function of Muchnik and Zavalishin 0 0

12 Synaptic weights having the shape of “Mexican hat” Weight Y X X Weight Weight X Y Y X X Information functions proposed by Muchnik and Zavalishin X 0 0 0 0 0 0

13 Scanning the image with an attention field of decreasing size Separating an object from background, measuring the object position and elongation parameters Adapting the position and shape of attention field Paraboloid of revolution Matching and masking the sub-image inside the AF

14 1. Mean local brightness. 2. Local spatial density of distribution of texels. 3. Average sizes (length and width) of texels. 4. Spatial orientation of texels. 5. Distance between neighboring similar texels connected with permissible straight line. 6. Orientation of permissible straight line connecting similar texels. Neuro-physiological texture description after David Marr Bela Yulesz surmised that there are special texel detectors in the living vision systems of the shape of “Mexican hat” proposed by David Marr. 14

15 Texture structural description by convolutional neural network of Lutsiv and C° Aerial photograph Result of image segmentation by texture features Texels presented as oriented elongated micro-objects Historgam of texel types Clustering of texture description feature vectors Feature vector х Weight Detectors of texel borders after Vistnes Weight Detectors of texels after Vistness Detectors of texels after Lutsiv 15

16 Training of Supervision’s deep convolutional neural network Examples of training images

17 Backward pass Training the Supervision’s deep convolutional neural network Backward propagation of error signal Too deep networks encounter difficulties in training by backward propagation of error signal

18 Pretraining of weights by restricted Boltzmann machine

19

20

21

22

23

24

25 Training acceleration by changing of sigmoid (logistic, activation) function

26

27 256 224 224 256 The

28 Training of Supervision’s deep convolutional neural network 48+48

29

30 Image classification in Supervision’s deep convolutional neural network max

31 Images classified as belonging to same classes

32

33 Is

34 Any questions?