Member-only story

Creating Convolutional Neural Networks from Scratch:

Victor Sim
Analytics Vidhya
Published in
5 min readAug 20, 2020
Photo by Zoltan Tasi on Unsplash

In recent years, Convolutional Neural Networks have been the first choice for Image processing in Machine Learning projects. Let’s look at how Convolutional Neural Networks work and try to create one for ourselves.

What are Convolutional Neural Networks?

Convolutional Neural Networks are an optimized type of neurons, that can extract features from a n-dimensional array.

Credit to:https://icecreamlabs.com/2018/08/19/3x3-convolution-filters%E2%80%8A-%E2%80%8Aa-popular-choice/

Here is how a Convolutional Neural Network functions:

Step 1: A filter is generated.

Step 2: The filter is applied on every part of the image, even in overlapped regions. (Look at animation above for visual aid).

Step 3: Multiply the filter weights by the region covered by the filter.

Step 4: Sum all the values together.

With these four steps, a full forward pass of a Convolutional layer is complete.

Going Through The Code:

Step 1| Dependencies:

import numpy as np
from matplotlib import pyplot as plt

--

--

Analytics Vidhya
Analytics Vidhya

Published in Analytics Vidhya

Analytics Vidhya is a community of Generative AI and Data Science professionals. We are building the next-gen data science ecosystem https://www.analyticsvidhya.com

Victor Sim
Victor Sim

Written by Victor Sim

Interested in Machine Learning. Open to internships and opportunities. Connect at https://linktr.ee/victorsi.

No responses yet