川崎のシステム開発・アプリ開発・Web制作

Convolution in Neural Network

1. Convolution

In the neural network, convolution is the name of the process to take image and get features from this image. Convolution comes to help condense the image down to the important features, some convolution will change the image in such a way that certain features in the image get emphasized. Like bellow , with different convolution we’ll get the different image with different features.

2. How convolution work

First, in every image, take every pixel and take its value and look at the value of its neighbor and to get the new value for the pixel, convolution simply multiplies pixel and each neighbor pixel by the corresponding value in the filter then get sum of this to get the new pixel with the new value.

3. Convolution in programing

  • In TensorFlow 2.0 we can use convolution by call tensorflow.keras.layers.Conv2D
  • In python :
where x_transformed and y_transformed is height and width of the image
この記事を書いた人

コメントを残す

メールアドレスが公開されることはありません。 * が付いている欄は必須項目です