in Blog

July 28, 2021

Machine Learning in Computer Vision


Edwin Lisowski

CSO & Co-Founder

Reading time:

5 minutes

On our blog, we talk about both these technologies frequently. In fact, they go hand in hand in many instances, working together to create improved, more effective, and accurate applications and algorithms that simply do their job much better. In this post, we want to take a closer look at how machine learning is used in computer vision services and how these two technologies overlap.

d you what there is to remember about CV and ML. In short, computer vision algorithms rely on various instruments, lenses, image sensors, and vision processing software. The main goal of this technology is to “see” the environment similar to the human eye and to interpret these images. This way, thanks to CV algorithms, modern cars can spot the pedestrian even before the driver does. But you have to understand that this technology is extensively used in many sectors, including manufacturing, healthcare, and even retail.

machine learning in computer vision
We also have a technology that’s called machine learning. It’s a subset of AI that allows algorithms, machines, and applications to learn themselves based on a) prior training and b) incoming data. A good way to illustrate how ML works is through chatbots. You see, advanced chatbots and voicebots are fueled with machine learning.

Thanks to this solution, they can learn from each conversation they have. After tens and hundreds of interactions, they can do their job much better and offer more accurate responses to diverse user queries, even those that haven’t been initially included in their training.

At this point, a vital question would be, how do these two technologies work together?

Machine learning in computer vision: Synergy

So we have two different disciplines of AI with different capabilities. As it turns out, though, these two technologies can work together, resulting in synergy. First off, we have to highlight that we have two primary stages in computer vision:

  • Sensing stage
  • Interpretation stage

“Seeing” is one thing. Understanding what you see and how to react to it is a completely different story. Now, machine learning in computer vision works best in the sensing stage. With diverse cameras, radars, and sensors, you can create a comprehensive image of what’s going on in the immediate surroundings of the car, machine, or device. However, you still have to teach your algorithm to interpret what it sees and react accordingly.

machine learning
Let’s use a simple example: Suppose you have a smart car assistant that detects pedestrians on crosswalks. Computer vision is responsible for monitoring car’s surroundings for crosswalks and pedestrians, but it’s the machine learning part that helps them detect each crosswalk and pedestrians on them. In other words, CV sees, and ML tells what CV sees. And when it comes to the interpretation stage, machine learning is indispensable.

It might be interesting for you: Computer Vision in Fintech

Augmented reality

AR is the perfect example to show you how machine learning in computer vision works. In order to build a decent AR app, for instance, for your online store, you need a computer vision part (so that the app can use your smartphone’s camera and capture images coming from it), and a machine learning part (for advanced image and scene labeling). Machine learning models built in retail AR apps help classify the location (by labeling scenes) and improve object detection (by estimating objects’ position and size).

Now, let’s find out how this cooperation works in the real world. Today, the vast majority of computer vision algorithms use machine learning, simply because this combination works brilliantly. And we have some interesting examples to showcase!

Lowe Bots

This American home improvement store uses intelligent customer assistants that gather real-time data by using machine learning algorithms in computer vision to scan inventory in a specific store and alert the store service when there’s a shortage of a given product or any other discrepancies.

Microdoft Inner Eye

This software can be used in hospitals and clinics to identify possible tumors and other anomalies found thanks to the X-ray images. The radiologists can upload the three-dimensional patient scans into the system. Next, the Inner Eye algorithms analyze the picture and mark areas that potentially contain tumors or other anomalies. Microsoft’s software uses both computer vision and deep learning (which is a more advanced version of machine learning).

If you want to find out more about this fascinating project, take a look at this video:

BMW Production lines

BMW extensively uses computer vision in their production plants to improve the production quality, verifying model designations, and detecting defective parts and elements to sort them out. Take a look at how this works in BMW factories:

To sum up, machine learning and computer vision and closely connected today, making both these technologies something better and more accurate. At Addepto, we provide machine learning consulting services. Drop us a line today, and see how we can help with your machine learning and computer vision project.


Machine Learning

Computer Vision