<div id="readability-page-1" class="page"><div>

<a href="https://blogger.googleusercontent.com/img/a/AVvXsEh6zmxX_P7YWryFnPgpjWx8p-z-h_2TmazphkBDsAuBC-zsBY8D_d-mOCC8NtiubS07m5ajL2zK1Yhm4WZwuagkXvq8q5nD6TfOyVhLT3kiUk10jMeal75DSLYfChd4zSclsbHiXGDdBcBuZoraIdzaMBnmQiR2UTZNRZsOmXKuquaXgZywXdIeEjqZ=s1596" imageanchor="1"><img data-original-height="1093" data-original-width="1596" src="https://blogger.googleusercontent.com/img/a/AVvXsEh6zmxX_P7YWryFnPgpjWx8p-z-h_2TmazphkBDsAuBC-zsBY8D_d-mOCC8NtiubS07m5ajL2zK1Yhm4WZwuagkXvq8q5nD6TfOyVhLT3kiUk10jMeal75DSLYfChd4zSclsbHiXGDdBcBuZoraIdzaMBnmQiR2UTZNRZsOmXKuquaXgZywXdIeEjqZ=s640" width="50%"></a>






Recently, deep learning is being
 applied to numerous areas and it shows amazing performance especially in the
 computer vision area. Computer vision is an area that interprets the vision
 information and utilizes it by computers. You can do image classification,
 object detection, optical character recognition, and many other fascinating
 computer vision tasks using deep learning.



 In this post, let's take a look at the
 Convolutional Neural Network&nbsp;(so-called CNN) which is used for
 deep learning computer vision tasks.&nbsp;




<img data-original-height="264" data-original-width="290" src="https://blogger.googleusercontent.com/img/a/AVvXsEiYD47FU-YJkx5LH668Z06eUOrnfW6Tj_ap4GyEecpgrpO4OuY-BFFKFLqYzK8EDAXJTJmZekNTs6atDYR77JjBxn0F5gTbcydM812b5rst5GpIv_hiOx_7j4lvTQoxGDmgmy6En700aA_R-yJ0Der7O7DxBk5BzCs-IqXIP-PqWtc9WR7sNLC32n-l=s1024" width="25%">


Let's say there is handwritten image data representing "2".

 This data is saved in the computer as a 2-dimensional array like below.



<img data-original-height="372" data-original-width="380" src="https://blogger.googleusercontent.com/img/a/AVvXsEgiJ8bD4vxxXPLsReRTaJ1MxDzEvj7u0f4Dpv-eMM_7m_V8om1BLTCS_wl3dgXasVMDaVYfK-m9mbxXmTSm0rdZxBaxgFrM2iYbx__ENFVOziPk9WlD0DgTXbGoCqziPs_QsBMLHF0ZJlehr_VzJS4GRx11qEYIJpQdFIRwvGdA9p-gTIqDywePF4ul=s1024" width="35%">



 When you say 10-megapixel resolution about your cell phone camera, it means
 there are 10 million cells in a 2-dimensional array on the picture taken by
 that camera (with maximum resolution setting.) This cell is called "pixel". So
 our example data is composed of 25 pixels (5 X 5).



 Each pixel has a value that represents the brightness of its location.


 In the case of black and white image data, 0 means black and 1 means
 white.&nbsp;


<img data-original-height="367" data-original-width="376" src="https://blogger.googleusercontent.com/img/a/AVvXsEjOq_Qgcjg_M3GhgvJnVKE9oNqqH_ixoaIfZ2-WwDdlxT1C6ySgu2h-sOx8YK2OwgEpdWEnVxXbmUM_ARZKkwq9ZPa2ifETj34ZJTxPkwv0-o0NeHAEj89oDtskX0bKo_PZzA8cnsAi-F_i3IAoa2HUATbHkBc3G0eYQrbWVBte0N8TkB9lEmFjvZUO=s1024" width="35%">



 By this logic, the data of "2" is a 2-dimensional array that has 25 values of
 brightness. To make it simple, I compose it with only four values - 0, 0.3,
 0.6, 1.


Now let's add some transformations here.

<img data-original-height="410" data-original-width="700" src="https://blogger.googleusercontent.com/img/a/AVvXsEhJFuRVo7OCl4ID57o2J6iOfLmUDqlOQ7gXfilYr11uvF8sWO9tneupKooi1ddwBRk_okjvq4OHnwpjNlU8SDwtfsNgjZKePqQubMdpl0VlpwpAagq4-SFBNY-00Ouu6uRP8QVS5GtXTTyQF16OU5IXjws5UK-oSvAWwB1tbS2k9GXIzBGHQCDyMJ2M=s1024" width="70%">

<div>
 
 We will transform the data on the left using the 3 X 3 size array on the
 right. This red array is called a filter.
</div>

Transformation is simple.&nbsp;
<div>
 The filter slides on the data, multiplying the values on the same pixel. You can understand this as a weighted sum operation - summing data
 values after multiplying each value's weight.&nbsp;
</div>




<a href="https://blogger.googleusercontent.com/img/a/AVvXsEjl2SmdDnubEOeZXo38hqC6jcFCyJvQa6ZdNh7EDYXKTs7a3JkcudzGGXUUf6Mk-5tWg8Kf4qZQ9oLBCHCyOt489OHTdVj9ak1mljjkqq3tgzxNjXuo5F9owvm3K1eJg08Cml-Ms3xOoCKj6papm4InizNvC5CJMTHQUN-NeroKzGIte_AUtHCm_vGZ=s1242"><img data-original-height="811" data-original-width="1242" src="https://blogger.googleusercontent.com/img/a/AVvXsEjl2SmdDnubEOeZXo38hqC6jcFCyJvQa6ZdNh7EDYXKTs7a3JkcudzGGXUUf6Mk-5tWg8Kf4qZQ9oLBCHCyOt489OHTdVj9ak1mljjkqq3tgzxNjXuo5F9owvm3K1eJg08Cml-Ms3xOoCKj6papm4InizNvC5CJMTHQUN-NeroKzGIte_AUtHCm_vGZ=s1024" width="75%"></a>


The result of the operation in the first slide is as above.
Let's go line by line.&nbsp;


 1. Multiplying the first row of the data&nbsp;[1,&nbsp; 0.6,&nbsp; 0.3]


 &nbsp; &nbsp; and the first row of the filter&nbsp;[0,&nbsp; 1,&nbsp; 0]


 &nbsp; &nbsp; results in
 [0,&nbsp; 0.6,&nbsp; 0]&nbsp; &nbsp;(1 x 0 = 0, 0.6 x 1 = 0.6, 0.3 x 0 = 0)



 2. Multiplying the second row of the data&nbsp;[1,&nbsp; &nbsp;0,&nbsp; &nbsp;0]


 &nbsp; &nbsp; and the second row of the filter&nbsp;[1,&nbsp; &nbsp; 2,&nbsp; &nbsp;1]


 &nbsp; &nbsp; results in&nbsp;[1,&nbsp; &nbsp;0,&nbsp; &nbsp;0]



 3. Multiplying the third row of the data&nbsp;[1,&nbsp; &nbsp; 1,&nbsp; &nbsp;0]


 &nbsp; &nbsp; and the first row of the filter&nbsp;[0,&nbsp; &nbsp;1,&nbsp; &nbsp;0]


 &nbsp; &nbsp; results in&nbsp;[0,&nbsp; &nbsp;1,&nbsp; &nbsp;0]



 4. Summing all of those values gives&nbsp;2.6. This is the result of this operation.


<div>
 <a href="https://blogger.googleusercontent.com/img/a/AVvXsEh9pYqTtJzstnI6VmBFguAUN56lJPjIUK9nA-zgBOLYa2u7r9AeghEAmeCSVHkIviax5TBs_un3uUyuWDOMbvTQcwaXmK9EBLzedudUnuBltAnWTYI8zpvVwkNQHLELXEOfXNZf2oJIJp_pJAXYcA7WLQAE51cNvHBXwKpj7LOCqVWdZrKKnT4ley_Z=s1295"><img data-original-height="759" data-original-width="1295" src="https://blogger.googleusercontent.com/img/a/AVvXsEh9pYqTtJzstnI6VmBFguAUN56lJPjIUK9nA-zgBOLYa2u7r9AeghEAmeCSVHkIviax5TBs_un3uUyuWDOMbvTQcwaXmK9EBLzedudUnuBltAnWTYI8zpvVwkNQHLELXEOfXNZf2oJIJp_pJAXYcA7WLQAE51cNvHBXwKpj7LOCqVWdZrKKnT4ley_Z=s1024" width="80%"></a>
 
 
</div>
The next operation starts after the filter slides one cell.

 The values in the filter are fixed, and they will be multiplied by the values
 in the same pixels of the data. Then, we save this second result next to the
 previous result.


<img data-original-height="389" data-original-width="627" src="https://blogger.googleusercontent.com/img/a/AVvXsEhTiQTnMb5xNg5IrWvvvKoaJowNFAJLoWaSBadd7lCd6AyIqa7i6UTKIwPojp3gvGk2vqTz62fnksUnE4jCApjepVNvsCKsMb3yIrFuTGRlQN7Y-EaDqUtS1V1Dkn10LkHoFqSHVaHJWXjcBKbfhF0J78KpMabBKDWoj9TuOBVKPBchLYxsnZDLOenU=s1024" width="60%">

And so on.
Now the filter reached the right end of the data. What can we do?

<img data-original-height="379" data-original-width="664" src="https://blogger.googleusercontent.com/img/a/AVvXsEgHcnK0gRMwg2YK3KJ3fqVF51SbISJxafRDR29Cua-PtTIXvbN9X_mPg9lSjFNkMnL1zcNQtCpRaFLWKlwWbYP1PQpO5p-opJdw9-iDPCJWY9uiLPkxKss7Okn2lY4fIiyxzzCjbuEo3IbdpkpOFPTBW1gotyfTOLEwv4ChFjezVx1XWktjVMU8yOY1=s1024" width="60%">


It goes one cell down and starts from the left again.

 When you finish this process until the end of the data, you will get the 3 X 3
 size result.



<a href="https://blogger.googleusercontent.com/img/a/AVvXsEhN4xi8-BPWYFgQyPWRH-9YGC1R0TKFUCKvhHaltcv2-0hRQP_QkBnU-ohZSPPxthmAyO4ebosmqG_DEzRyGjU3_3hFg0lgeAqNhKsXO9Z-L_Dv9m29nEvHiGWunqp5NVHZ5j36_7dwm1QKNSt31O3fVaN8YLkO02xxqgIwJfDmjUXzk8r0unDHVl9i=s1257"><img data-original-height="853" data-original-width="1257" src="https://blogger.googleusercontent.com/img/a/AVvXsEhN4xi8-BPWYFgQyPWRH-9YGC1R0TKFUCKvhHaltcv2-0hRQP_QkBnU-ohZSPPxthmAyO4ebosmqG_DEzRyGjU3_3hFg0lgeAqNhKsXO9Z-L_Dv9m29nEvHiGWunqp5NVHZ5j36_7dwm1QKNSt31O3fVaN8YLkO02xxqgIwJfDmjUXzk8r0unDHVl9i=s1024" width="75%"></a>


Why 3 X 3?

 Let's think about the width only. The length of the data is 5 and the length
 of the filter is 3. So filter does the operation 3 times.* Same as height, so
 the number of operations becomes 3 X 3, 9.


* If filter slides 2 cells each time, then filter locates only two times.
 So the size of the result will be 2 X 2, 4. This parameter that determines
 how many cells the filter slides each time is called stride. If you
 want to make the size of the result as same as the original data, then you
 can surround the data with zeros before starting the sliding operations.
 Then the size of the data will be 7 X 7, and the size of the result will be
 5 X 5. This trick is called (zero)&nbsp;padding.



This sliding operation is called convolution.*

* This does not have the exact same meaning of
 <a href="https://en.wikipedia.org/wiki/Convolution">convolution</a> in
 mathematics, but we use it this way in the deep learning field.

<div>
 
 <a href="https://blogger.googleusercontent.com/img/a/AVvXsEgDz4ti5KbmCm0dY__FL7LG8ynaHr5VB8oNalzmFt_0JpZCeDlPv8izWaYpFO3OQrDbV2nUyuQN5pZWNiWak-b20u6pKoAdWLawcs2hY5sPOLFDqs_wBHmtoHEHINBVtEumLOoZHWvJ3lDCCeeBNtAizv_gCXmSP9vhy7tjbR_l3lhCAHnI3tqCg8mf=s1272"><img data-original-height="862" data-original-width="1272" src="https://blogger.googleusercontent.com/img/a/AVvXsEgDz4ti5KbmCm0dY__FL7LG8ynaHr5VB8oNalzmFt_0JpZCeDlPv8izWaYpFO3OQrDbV2nUyuQN5pZWNiWak-b20u6pKoAdWLawcs2hY5sPOLFDqs_wBHmtoHEHINBVtEumLOoZHWvJ3lDCCeeBNtAizv_gCXmSP9vhy7tjbR_l3lhCAHnI3tqCg8mf=s1024" width="75%"></a>
 
</div>


 Now let's do the same process using another filter. Since the filter has
 different values this time, the result will be different.


<a href="https://blogger.googleusercontent.com/img/a/AVvXsEgrUEVy3dnAzZnQgd0CQxxhr-7T3ZgN7ZYJQYPL0KA5Ha_N2r_HiUthwSTuNjSfo_5mTpKPkIxtrLNKl-BbRfJKZRuzFUSLWpm1ZmrD8EcEQydDGLRhQUhIc4m6w0F4oBCXoh7dfu_L95M3DfMUV2PZFQmqzN0shr54NqM4qfuxLHJgj3VYXMcysm3C=s1397"><img data-original-height="895" data-original-width="1397" src="https://blogger.googleusercontent.com/img/a/AVvXsEgrUEVy3dnAzZnQgd0CQxxhr-7T3ZgN7ZYJQYPL0KA5Ha_N2r_HiUthwSTuNjSfo_5mTpKPkIxtrLNKl-BbRfJKZRuzFUSLWpm1ZmrD8EcEQydDGLRhQUhIc4m6w0F4oBCXoh7dfu_L95M3DfMUV2PZFQmqzN0shr54NqM4qfuxLHJgj3VYXMcysm3C=s1024" width="75%"></a>



 We can keep doing this by using various filters. We will have multiple result
 arrays of the convolution operations. So what's the meaning of doing this?


 The first cell in a result array comes from multiplying the left-top 3 X 3
 part of the data and the filter. This means a filter summarized 3 X 3
 cells into 1 cell by the convolution operation.


Doing the convolution operations using filters, therefore, generates multiple
results which have summarized and interpreted regional information of the
original image data. Each pixel of the result arrays contains 9-pixel
information - meaning that the computer can see multiple pixel information at
the same time with one value.

<div>
 In other words, the result arrays are the extracted features from the
 original image data. If you use 5 filters, then you will get 5 new features
 from the data. Now we get the familiar structure - Neural Networks.<a href="https://blogger.googleusercontent.com/img/a/AVvXsEgjhrpqIENkl5pw3W-jy1Pjzgz0Nt-1QbctegSzX9yhaHQpH8OkZoM-Z4xIyIMT8ti2s159Tvwms0qUDLNYdH6xM300lGsQP9YksKFZRDntFdPhTG37VTu9XbpRg1ZiHvoSLEI_R_UFaEwtz_SZrkefoYeuMx_bcDuX9_jLZz5fcO7o6UNKWtlD49YM=s1209"><img data-original-height="909" data-original-width="1209" src="https://blogger.googleusercontent.com/img/a/AVvXsEgjhrpqIENkl5pw3W-jy1Pjzgz0Nt-1QbctegSzX9yhaHQpH8OkZoM-Z4xIyIMT8ti2s159Tvwms0qUDLNYdH6xM300lGsQP9YksKFZRDntFdPhTG37VTu9XbpRg1ZiHvoSLEI_R_UFaEwtz_SZrkefoYeuMx_bcDuX9_jLZz5fcO7o6UNKWtlD49YM=s1024" width="90%"></a>
 
 
 
 Result array from a convolution operation using a filter is called a&nbsp;feature map. The number of feature maps is called the number of channels as
 well. If you use 10 filters, then there will be 10 output channels.
 
 
 
 Having multiple channels means that there are multiple 2-dimensional arrays,
 which is identical to having one 3-dimensional array. We used single channel
 (black and white) data, but the colored image data has 3 channels - Red,
 Green, Blue channels.
 
 
 <div>
 <a href="https://blogger.googleusercontent.com/img/a/AVvXsEho0xfsra3w9bnSm5HisIoyMkW6PxtWI8iVX77MTIgWc_SDtoX-ZOLx1puy-npFiyqsBCN6hk_VshsuJl6U8cXLkeH3Ud_jF2sbrhpmJGtnnCi2DEhpGArjJBBD3uXFo5kVJfyAfYCcCO7oYD6SYSlIcTjOU0A7IZflzaW85e_cvi6WzjmsdbtAnbTd=s1420"><img data-original-height="668" data-original-width="1420" src="https://blogger.googleusercontent.com/img/a/AVvXsEho0xfsra3w9bnSm5HisIoyMkW6PxtWI8iVX77MTIgWc_SDtoX-ZOLx1puy-npFiyqsBCN6hk_VshsuJl6U8cXLkeH3Ud_jF2sbrhpmJGtnnCi2DEhpGArjJBBD3uXFo5kVJfyAfYCcCO7oYD6SYSlIcTjOU0A7IZflzaW85e_cvi6WzjmsdbtAnbTd=s1024" width="85%"></a>
 
 
 </div>
 So what kind of values should be in the filters?
 
 Before the era of deep learning, the values in the filter were fixed. Have
 you ever used "shape filter" in a photo editor? The red filter we used
 earlier is actually almost the same as the shapen filter. (Try converting 1s
 to -1)
 
 
 <div>
 However, the values are not fixed when it comes to deep learning.
 The key principle of learning is changing the values of the weights into
 certain directions so that the model makes better predictions.
 The values in a convolution filter are the weights of a Convolutional
 Neural Network.
 </div>
 
 <a href="https://blogger.googleusercontent.com/img/a/AVvXsEjPy2pBfqOUYPLXbCU8yPOH-DBwshFOTMKE6pqNzWbDzC7clKiQJRdppBoLEBHBJjnlNiFplCghNB_-E6h7u8KNWgagLS-MW3Q6eq6WuK95Pr09ysI7xjmoFk7shyqCKOsBA1dWcqfs4lM9ZeJi6M_pr49fVY-niIqtC1_TpnFg21csrKiKtjtm9-yM=s1065"><img data-original-height="662" data-original-width="1065" src="https://blogger.googleusercontent.com/img/a/AVvXsEjPy2pBfqOUYPLXbCU8yPOH-DBwshFOTMKE6pqNzWbDzC7clKiQJRdppBoLEBHBJjnlNiFplCghNB_-E6h7u8KNWgagLS-MW3Q6eq6WuK95Pr09ysI7xjmoFk7shyqCKOsBA1dWcqfs4lM9ZeJi6M_pr49fVY-niIqtC1_TpnFg21csrKiKtjtm9-yM=s1024" width="65%"></a>
 
 
 
 That means, we put some random values into the filters and set how many
 filters we use in each layer before starting training, but the values in the
 filters change as the training goes on. By doing so, the values in the
 filters are getting optimized, more meaningful feature maps are being
 extracted, so the model makes better predictions.
 
 <a href="https://blogger.googleusercontent.com/img/a/AVvXsEgn3COVR1HtHpqIhpse5JMAhhS1PvBsbMBldkqjmyvBSZWv1pRzmbqHz2qscTAt9WlcHe3zve7fsEcQtGB8Bkl47NpxdyB3vjzkhowNsJwZeq1WC-9DdNKNcX-XTn_U6mvS6OA8KtUjDOrsEZKYLTrLICqTjdyPg76Tb8kUZFYWlQyDqW7PFA6DhJnR=s1431"><img data-original-height="1039" data-original-width="1431" src="https://blogger.googleusercontent.com/img/a/AVvXsEgn3COVR1HtHpqIhpse5JMAhhS1PvBsbMBldkqjmyvBSZWv1pRzmbqHz2qscTAt9WlcHe3zve7fsEcQtGB8Bkl47NpxdyB3vjzkhowNsJwZeq1WC-9DdNKNcX-XTn_U6mvS6OA8KtUjDOrsEZKYLTrLICqTjdyPg76Tb8kUZFYWlQyDqW7PFA6DhJnR=s1024" width="95%"></a>
 
 
 See, Convolutional Neural Network is also just a Neural Network.
 
 
 
 
 
 
 <img data-original-height="1252" data-original-width="1670" src="https://1.bp.blogspot.com/-OD6Gtx0AyaY/X1kN_knmsEI/AAAAAAAAkrY/D-dYGyZDf6QmHfUkkzlcFpjMMCqp71i5gCLcBGAsYHQ/s960/%25EC%2588%2598%25EA%25B3%25A0%25ED%2596%2588%25EB%258B%25A4%25ED%259C%25B4%25EB%25A8%25BCen43.png" width="320">
 
 
 
 
 
 Reference for going deeper
 
 <div>
 <ul>
 <li>
 &nbsp;<a href="https://setosa.io/ev/image-kernels/">Image Kernels Explained Visually</a>
 </li>
 <li>
 &nbsp;<a href="https://www.cs.ryerson.ca/~aharley/vis/conv/">3D Visualization of CNN</a>
 </li>
 <li>
 &nbsp;Stanford&nbsp;CS231n [<a href="https://youtu.be/bNb2fEVKeEo">lecture</a>]&nbsp;[<a href="https://cs231n.github.io/convolutional-networks/">note</a>]
 </li>
 </ul>
 </div>
 
</div>
</div></div>

국문과 공대생

[한글] Read this in Korean
  Recently, deep learning is being   applied to numerous areas and it shows amazing performance especially in the   computer vision area. Computer vision is an area that interprets the vision   information and utilizes it by computers. You can do image classification,   object detection, optical character recognition, and many other fascinating   computer vision tasks using deep learning. 
  In this post, let's take a look at the   Convolutional Neural Network (so-called CNN) which is used for   deep learning computer vision tasks.  
  
Let's say there is handwritten image data representing "2".
  This data is saved in the computer as a 2-dimensional array like below. 
  
  When you say 10-megapixel resolution about your cell phone camera, it means   there are 10 million cells in a 2-dimensional array on the picture taken by   that camera (with maximum resolution setting.) This cell is called "pixel". So   our example data is composed of 25 pixels (5 X 5). 
  Each pixel has a value that represents the brightness of its location. 
  In the case of black and white image data, 0 means black and 1 means   white.  
  
  By this logic, the data of "2" is a 2-dimensional array that has 25 values of   brightness. To make it simple, I compose it with only four values - 0, 0.3,   0.6, 1. 
Now let's add some transformations here.
  
  
  We will transform the data on the left using the 3 X 3 size array on the   right. This red array is called a filter. 
Transformation is simple. 
  The filter slides on the data, multiplying the values on the same pixel. You can understand this as a weighted sum operation - summing data   values after multiplying each value's weight.  
  
The result of the operation in the first slide is as above.
Let's go line by line. 
  1. Multiplying the first row of the data [1,  0.6,  0.3]
      and the first row of the filter [0,  1,  0]
      results in   [0,  0.6,  0]   (1 x 0 = 0, 0.6 x 1 = 0.6, 0.3 x 0 = 0) 
  2. Multiplying the second row of the data [1,   0,   0]
      and the second row of the filter [1,    2,   1]
      results in [1,   0,   0]
  
  3. Multiplying the third row of the data [1,    1,   0]
      and the first row of the filter [0,   1,   0]
      results in [0,   1,   0]
  4. Summing all of those values gives 2.6. This is the result of this operation. 
  
      
  
The next operation starts after the filter slides one cell.
  The values in the filter are fixed, and they will be multiplied by the values   in the same pixels of the data. Then, we save this second result next to the   previous result. 
  
And so on.
Now the filter reached the right end of the data. What can we do?
  
It goes one cell down and starts from the left again.
  When you finish this process until the end of the data, you will get the 3 X 3   size result. 
  
Why 3 X 3?
  Let's think about the width only. The length of the data is 5 and the length   of the filter is 3. So filter does the operation 3 times.* Same as height, so   the number of operations becomes 3 X 3, 9. 
  * If filter slides 2 cells each time, then filter locates only two times.     So the size of the result will be 2 X 2, 4. This parameter that determines     how many cells the filter slides each time is called stride. If you     want to make the size of the result as same as the original data, then you     can surround the data with zeros before starting the sliding operations.     Then the size of the data will be 7 X 7, and the size of the result will be     5 X 5. This trick is called (zero) padding.
This sliding operation is called convolution.*
  * This does not have the exact same meaning of     convolution in     mathematics, but we use it this way in the deep learning field.
  
  
      
  Now let's do the same process using another filter. Since the filter has   different values this time, the result will be different. 
  
  We can keep doing this by using various filters. We will have multiple result   arrays of the convolution operations. So what's the meaning of doing this? 
  The first cell in a result array comes from multiplying the left-top 3 X 3   part of the data and the filter. This means a filter summarized 3 X 3   cells into 1 cell by the convolution operation. 
Doing the convolution operations using filters, therefore, generates multiple results which have summarized and interpreted regional information of the original image data. Each pixel of the result arrays contains 9-pixel information - meaning that the computer can see multiple pixel information at the same time with one value. 
  In other words, the result arrays are the extracted features from the   original image data. If you use 5 filters, then you will get 5 new features   from the data. Now we get the familiar structure - Neural Networks.
  
      
  
  
    Result array from a convolution operation using a filter is called a feature map. The number of feature maps is called the number of channels as     well. If you use 10 filters, then there will be 10 output channels.   
  
  
    Having multiple channels means that there are multiple 2-dimensional arrays,     which is identical to having one 3-dimensional array. We used single channel     (black and white) data, but the colored image data has 3 channels - Red,     Green, Blue channels.   
  
  
    
          
    
  
  
So what kind of values should be in the filters?
  
    Before the era of deep learning, the values in the filter were fixed. Have     you ever used "shape filter" in a photo editor? The red filter we used     earlier is actually almost the same as the shapen filter. (Try converting 1s     to -1)   
  
  
    However, the values are not fixed when it comes to deep learning.     The key principle of learning is changing the values of the weights into       certain directions so that the model makes better predictions.    The values in a convolution filter are the weights of a Convolutional     Neural Network.   
  
  
      
  
  
    That means, we put some random values into the filters and set how many     filters we use in each layer before starting training, but the values in the     filters change as the training goes on. By doing so, the values in the     filters are getting optimized, more meaningful feature maps are being     extracted, so the model makes better predictions.   
  
  
      
  
  
See, Convolutional Neural Network is also just a Neural Network.
  
  
  
  
  
  
  
      
  
  
  
  
  
    Reference for going deeper  
  
    
      
         Image Kernels Explained Visually      
      
         3D Visualization of CNN      
      
         Stanford CS231n [lecture] [note]

Convolutional Neural Network Explained SO EASY

댓글