You Need to Update Your CV Knowledge
Here’s why your CV knowledge should be updated.
And no, we don’t mean your Résumé. Put that aside for now.
What is CV?
CV, or more popularly known as Computer Vision, refers to when a computer is being taught how to “see” images and pictures for what they are.
Now, does that mean computers have eyes just like us?
Well, not exactly. By see, we don’t necessarily mean they have the power of sight the way humans do. We mean something more along the lines of interpretation.
These computers don’t automatically know that whatever they are detecting is an image of an object or a person. They can only interpret and differentiate it all through codes. These codes are the special languages that help them function properly.
It’s like being born blind. The blind person has never seen what the world and all its properties look like. He can only feel around and touch things, but can’t exactly know what he’s touching, unless he guesses and pairs that with what he has been told that it is before.
It’s the same for computers, albeit with the difference that these computers are not human.
But I’m sure the question on your mind now is how ever in the world can they do that.
Let’s find out together.
How it all works.
It’s pretty straightforward. Well, the not so complex parts of Computer Vision, anyway.
In order to search or identify images, the underlying algorithm needs to first know what the images contain. It needs to know what properties exist in the image displayed before it, which is the core goal of Computer Vision.
For example, if we asked you to drop comments on what you can see in this picture below, you’d probably type out stuff like “Calculator”, “Jotter/notepad”, “Laptop”, and so on.
Now, it’s very easy for you to do so, identifying the things around you as you take a walk or a drive around your street. But you didn’t just do that on your own. Behind the scenes, there’s a series of functions that took place for you to be able to identify these objects.
Your eyes and your mental features all went through a form of communication, in order for you to know that what you are indeed seeing is a calculator and a jotter.
This is different in computers. To them, what they see is just a bunch of jumbled pixels. They don’t exactly know that there’s a laptop in that picture. Or a calculator. Or a jotter.
This is where CV comes in. This is the point where the human eye complexity has to be infused into that machine to replicate an almost human-like “seeing” performance or behaviour.
How do you use computer vision?
There are quite a lot of uses for Computer Vision, from industrial productions to image retrieval.
And the even more exciting part of this is that, with the looming increase in the number of self-driving cars, we should expect that we’ll be seeing more of Computer Visions, with time.
While you’re sitting back and relaxing, your car is doing most of the driving work. Computer Vision will prevent your car from running that red traffic light and end up running someone over, in turn.
All by detecting traffic lights on sight. By recognizing human faces and knowing when to slow down.
You even have a bit of Computer Vision present in your cell phones already. The face recognition feature that gives only you the power to unlock your cell is part of a larger Computer Vision function.
It can also identify images with objects in them, in order to properly arrange and categorize them. So, you don’t have to stress much on moving stuff to a particular group or folder.
All “food” related pictures will be moved to one group, and all “Selfies” with friends will be in their own separate groups, as well.
And if you have ever wondered why that picture of you on the beach in a skimpy bikini outfit was taken down from social media, it’s because Facebook or Instagram has incorporated a CV-based algorithm for their content moderation.
So, if there’s any hint of nudity or pornography, you can kiss your content goodbye.
And for all this to be achieved, a computer indirectly has to answer questions like “what am I supposed to be looking at here?”, Whenever it’s faced with a recognition task.
What objects are in this picture? Where is the object in the picture? Is it in front of a tree? What segment of the pixels in the image belongs to the object?
Challenges in computer vision.
Perhaps, the main challenge in Computer Vision is that computers are NOT human. It is obvious as it is true. Computers can be quite constrained or limited in their functions and understanding of our world.
You can teach a computer to recognize that 1+1=2, but to teach it to behave exactly like a human being? That’s an entirely different thing. Forget how the movies may make it seem easily possible.
We have tried to emulate this in cameras and succeeded, to some extent, but it’s different this time, because the new goal here is to replicate the human eye, which if you didn’t already know, is the most flexible “camera” there is, and arguably the best.
And to replicate such a function into machines built on codes, rather than made up of biological components like humans, is one major challenge for Computer Vision.
The best these machines can do is to compare pixels together to recognize them as different particles. They don’t know like us to differentiate parts of a thing and say “Okay, this part of the jotter is black and called the spiral bind. This is a Canon Camera. This is a blue Novel book”
So until you “teach” or train the machine’s algorithm to recognize these objects for what they are, as specifically unique existing materials, they won’t be able to put two and two together.
Who knows. Maybe Artificial Intelligence studies and research will make a significant breakthrough, in the next coming years, and there will be every possibility that machines can be as flexible as humanity is.
But until then, we are stuck with these suboptimal working metal versions of ourselves.