Original subtitles

Convolutional neural networks of CNN is the type of neural networks that uses convolution operation

in the process.

Deep learning can be used for solving computer vision problems, thanks to CNN.

Convolutional layer is the core layer of deep neural networks for solving computer vision problems.

We will dig deeper to understand the convolution process.

There are three major components in convolutional layer which are input data filters or kernels and

filter maps.

For example, suppose the input is a colour image with three dimensions which are width, height and

channel.

The filter of Corona is commonly referred to as a feature detector because it moves across the image

and performs a joint operation between the input and the value of the filter.

To produce an output which is a feature map.

Convolution is the name given to this process.

The filter is a two dimensional array with weights that can be updated during the training process.

By performing convolution into the image, we will have new image which contains features.

So how exactly convolution works?

This slide gives a simple illustration.

Given an image on the left and the filter of Col at the top.

Convolution is carried out by performing the product.

The filter to every part of an image starting from top left to the bottom right corner.

For example, from the first convolution operation, we will have a new value replacing the middle value

of the top corner, which is five into four.

We will have zero multiplied by four plus minus one multiplied by two plus zero multiplied by three

plus minus one multiplied by eight plus four multiplied by five plus minus one multiplied by two plus

zero multiplied by four plus minus one multiplied by four plus zero multiplied by three, which is equal

to four.

In this example, we use threat equals one, which means we slide the filter to the right by one pixel

and perform similar operation and we will have minus nine as the result.

Continue the process and we will have this feature map.

After creating the initial feature map, we typically perform a series of convolutions using various

filters, stride and padding to have various feature maps.

Later on, the networks will give large weights to feature maps that fit the label dataset and small

weights to the otherwise.

The number of filters has an impact on the output channel, for example, through different filters

to produce two different filter maps resulting in a two channel output.

The is the amount of filter displacement that occurs during convolution convolution with thread.

One is demonstrated in this and the previous examples if we use try to the mislead the filter to pixels

to the right and so on.

It is important to note that if we use try to the resulting feature map size will be even smaller.

This technique is used for DOWNSAMPLING, the feature map in YOLO fee for.

Pending is a process of ending border of close to the input limits.

There are three types of pending.

The first is without pending or so called valid PD.

As previously illustrated, the fiscal map generated by the convolution process will be smaller than

the input if we use valid PD.

The second is the same paddy.

The input image is ended with one layer of cross border.

The size of the feature generated by the convolution process will be the same as the input.

The third is the complete paddy the size of the Fitzroy map generated by the convolution process will

be larger than the input with this paddy.

For example, this is the input and this is the output.

The pooling layer in most deep neural networks is responsible for DOWNSAMPLING or reducing the feature

map dimension and a number of parameters.

This is important to speed up the process while maintaining important features.

Falling is commonly used in two ways maximum pulling or average pulling.

The maximum value of images covered by the kernel is to turn by max polling.

While the average value of image is covered by the kernel is returned by average poorly.

Please bear in mind that in this example we used try because to which means we slide it two by two pruning

kernel two pixels to the right and then to pixels to below right before the end of a deep neural networks

is a fully connected layer.

Because the feature map produced by the convolution and in layers is still a multidimensional array.

It must be reshaped into a vector before it can be used as input to the fully connected layer.

The reset process called flattening.

Fully connected layer is commonly used in multilayer perceptron applications and aims to transform the

data dimension so that the data can be classified linearly.