# Understanding Tensors

**00:00**
Welcome to this lesson where you’ll learn about tensors, which are the backbone of TensorFlow and PyTorch.

**00:08**
Think of tensors as containers. Imagine you have a container, and this container can hold a bunch of numbers like 3, 5, 8, and seven, and that’s a tensor.

**00:20**
It’s like a container that can hold numbers,

**00:23**
but here’s where it gets interesting. Tensors can hold multidimensional arrays. Let’s look at this picture. Tensors can be like rows of boxes or containers where each of them represents a piece of data.

**00:36**
Machine learning models use these rows of boxes to process information.

**00:41**
For example, 3D tensors can represent more complex data like images. How you might ask, let’s create a three by three tensor for a simple pixelated grayscale image.

**00:54**
Each cell in the tensor corresponds to a pixel with white and black colors, representing the maximum and minimum values respectively. Okay, here’s the tensor that you’re creating with an `np.array`

.

**01:07**
In the first row, you have `255, 0,`

`255, 0`

is a pixel that corresponds to black, and `255`

corresponds to white, and in the second row you have `0, 255, 0`

or black, white, black.

**01:22**
And on the third row you have `255, 0, 255`

again. Now, what would this correspond to as a whole? Let’s see.

**01:31**
Here you go exactly how you described it. In the tensor, you have a grayscale image with each row corresponding to the black and white pixels.

**01:40**
In a nutshell, tensors are just containers for numbers, and they come in different shapes and sizes depending on the data you’re working with.

Become a Member to join the conversation.