A Comprehensive Guide to Convolutional Neural Networks | by Maryam Sikander | Jul, 2024

On this article, we’ll discover Convolutional Neural Networks(CNNs), the basic structure of CNNs, and perceive how they work.

So Let’s Roll!!!

Convolutional Neural networks(ConvNets or CNN) are a kind of neural community impressed by how our mind processes visible data. CNNs are particularly designed for processing and analyzing visible knowledge, reminiscent of photographs and movies. Some areas the place CNNs are broadly used are picture segmentation, object detection, and face recognition.

The fundamental Structure of CNNs includes of three layers:

1. Convolutional layer
2. Pooling layer
3. Absolutely-connected layer

Let’s delve into every of those layers to get a deeper understanding of their working:

The convolutional layer is a elementary part of CNNs. It’s accountable for the convolutional operation, which entails making use of a set of filters (i.e. Kernels) to the enter knowledge to extract patterns and options. This layer helps in figuring out options, reminiscent of edges, textures, and shapes. This layer makes it simpler for the neural community to acknowledge advanced constructions inside photographs and movies.

These options are then mixed to create a function map which is then handed to the subsequent layer for additional preprocessing.

Key parts of the Convolutional layer:

→ Enter knowledge

→ Filters

→ Characteristic map

Convolutional Operation:

• The convolution operation entails sliding a filter (i.e. kernel) over the enter picture, computing the dot product between the filter and the enter, after which summing the outcomes to supply a brand new function map.
• This operation produces a brand new array known as the function map or activation map.
• This course of is repeated for a number of filters, leading to a number of function maps, every capturing completely different options of the picture.

Let’s stroll via an instance of how a filter(kernel) produces a function map:

First Place (Prime-Left Nook)

Overlay the filter on the top-left nook of the picture:

# Enter Picture                 Filter
1 2 3                          1 0
4 5 6                          0 1
7 8 9

# multiply and sum:
(1*1) + (2*0) + (4*0) + (5*1) = 1 + 0 + 0 + 5 = 6

So, the primary quantity in our function map is 6.

Transfer the filter one sq. to the best:

1 2 3      1 0
4 5 6  ->  0 1
7 8 9

(2*1) + (3*0) + (5*0) + (6*1) = 2 + 0 + 0 + 6 = 8

So, the subsequent quantity in our function map is 8

Repeat this for the whole picture.

The habits of the convolutional layer relies on the next hyperparameters: