A simple user-friendly calculator app to help you learn Convolutional Neural Networks with ease. Ace your exams and take your deep learning to the next level! This app teaches the fundamentals required for learning CNNs, and the math behind them. It helps you to map the input dimension of the Convolution to the output based on various hyperparameters set by the user and play with images.

default-2023

<pre class="ql-syntax" spellcheck="false">Try out an example!
</pre>

i_24

i_22

i_25

i_31

i_32

i_33

i_34

i_35

i_41

i_42

<h2>INPUT DATA﻿</h2>

i_43

i_44

i_45

i_51

i_52

i_53

i_54

i_55

<h2>KERNEL</h2>

k_11

i_11

k_12

k_13

k_21

k_22

k_23

k_31

k_32

k_33

The Convoluted Feature Map dimensions can be calculated by the given formula: <img src="https://miro.medium.com/v2/resize:fit:413/1*V5ZIZg7cGHLASKbnRbKBJQ.png" style="display: block; margin: auto;"> <h3>But what are padding and stride?</h3>Stride denotes how many steps we are moving in each step in convolution. Let's take a look at another example for stride=1: <img src="https://miro.medium.com/v2/resize:fit:875/1*g0OmDI1w9KqN7Rpw6Qo8Xg@2x.gif" width="432" style="display: block; margin: auto;">Similarly, we can have a stride of 2 or 3 inorder to jump 2 or 3 steps respectively. We can observe that the size of output is smaller than input. To maintain the dimension of output as in input, we use padding. Padding is a process of adding zeros to the input matrix symmetrically. In the following example, the extra grey blocks denote the padding. It is used to make the dimension of the output same as the input.<img src="https://miro.medium.com/v2/resize:fit:875/1*17TNPi4m0pBqOCGrXzU27w.gif" width="540" style="display: block; margin: auto;"><h3> </h3><h3>Now Let's try adding a stride and padding parameter to our previous example:</h3>

padding_1

i_13

Similarly, in the case of RGB color, channel take a look at this animation to understand its working: <img src="https://editor.analyticsvidhya.com/uploads/556091_ciDgQEjViWLnCbmX-EeSrA.gif" alt="In the case of an RGB color, channel take a look at this animation to understand its working" width="665" style="display: block; margin: auto;"> We can now work with images. Each image can be converted to an RGB matrix where each cell corresponds to the pixel value at the location. Many operations can be done by choosing an appropriate kernel.

PADDING:

STRIDE:

stride_1

image

Add your Image:

kernelchoice

<h3><a href="https://ibb.co/jwxJDJM" rel="noopener noreferrer" target="_blank" style="color: rgb(0, 102, 204); background-color: rgb(255, 235, 204);">Sample Image</a></h3>

Learn Convolutions

Similar apps: