The technology behind facial recognition, text and speech recognition, spam filters on email, online shopping and movie recommendations.
From medical diagnosis to social media, it’s all Machine Learning.
We encounter machine learning every day in our day to day lives. So it’s only fair we understand what it is and how it actually works.
What is Machine Learning?
Machine learning is the algorithm that allows computer programs to automatically improve through experience. It is one of the branches of Artificial Intelligence.
How does Machine Learning work?
Let’s start with a very well known example of cats and dogs.
If we need the computer to tell the difference between a picture of a dog and a picture of a cat, we begin by feeding it images and telling it this one is a dog, that one is a cat.
The computer programs/algorithms may then figure out statistical differences between dog and cat-like cats are generally shorter than dogs on their own. And then represent that information numerically.
Of course, there will be mistakes but the more data the algorithm receives the more finely tuned it becomes. And accurate in its predictions
The potential of machine learning to transform our world is truly amazing. In the video, we consider one such example of monitoring and counting the African Elephants, one of the vulnerable species to understand Machine Learning in action.
2:37 – How many days would it take to count all the elephants in a forest?
4:08 – Forming team to accomplish the elephant counting mission.
5:50 – How to count all the elephants on earth?
6:02 – A kid’s solution – flying planes with cameras and chips that count the elephants for us.
8:10 – The kid wasn’t far away from the solution. As part of an international project, researchers are using satellite images processed with computer algorithms, which are trained with more than 1,000 images of elephants to help spot them.
14:40 – Counting Elephants with Satellites – Can you spot any elephants in the image below? If yes, count them.
Similarly, a machine can be trained with thousands of images.
19:06 – Machine identified all the elephants
21: 10 – What can be the challenges in training a machine to recognize and count the number of elephants?
- Statues of elephants – machines can confuse the statues as real elephants.
- Newborns – Machine may not consider pregnant elephants which may give birth while we process the images already clicked.
- Shadows of elephants – Shadows may seem like real elephants.
- Cloudy Weather – Difficult to detect elephants due to low light.
- Dense forests – Elephants may be hidden behind shrubs and trees and cannot be seen by the satellite camera.
23:45 – Elephants are not the only animals that are being helped by Machine Learning, there are many. We have discussed one such program of saving Orca using machine learning.
What other challenges do you think could be faced while counting the elephants the above way? Do let us know in the comments below.
Here are a few more images for you to spot and count the animals in:
Siya Janakkumar Nimavat
at 3:28 pm
If there are shawdows of elephants means an elephant is also there , so how can it be wrong ?
at 7:12 am
It is not completely wrong. It just means that the machine/algorithm counts the same elephant twice, once the real elephant and then its shadow, which messes up the total count.