CODEWITHSUNDEEP
image-processinginputneural-networkperceptron
user1194497
user1194497

Reputation: 45

What exactly does an "input" refer in the case of a neural network doing image detection (for example)?

a simple neural network

Say, we have input in the form of a collection of images: - (200,56x56,3) where 200 is the number of distinct images, 56x56 are the pixels (length vs breadth) and 3 refer to RGB values

So, x1,x2,x3,x4 etc refer to (number of instances, pixels (length), pixels (breadth) and RGB value?

or are there 1,881,600 inputs (equal to 200x56x56x3)?

Upvotes: 0

Views: 130

Answers (1)

Antoine Dubuis
Antoine Dubuis

Reputation: 5314

The number of inputs in your case is 1*56*56*3=9408. Imagine that you want to predict a value for a 1 new image of dimension 56*56, you will have to feed the network with all RGB values (3) of every pixel.

In practice, feed-forward neural networks, as described in your picture, are not used for image classification. Instead, we are using CNN (Convolutional Neural Network).

Upvotes: 1

Related Questions