Exploring Data Science: Computer Vision and Beyond (Interview)

Today we’ll dive into the world of data science, where Michał, our Data Scientist, will shed light on his experiences with the field of computer vision.

Michał, can you give us a brief introduction about yourself, your background, and what drew you to the field of data science?

My name is Michał Myller; I am a Silesian University graduate with a Master’s in Data Science. Machine learning got me hooked from the get-go during my studies. Since then, I’ve tackled projects in classical machine learning, computer vision, and time-series analysis in healthcare, space, and industry over the last five years of my professional career.

At Addepto, I’m working as a Data Scientist, where I use my skills to bring customers’ ideas to life and elevate their products to new heights.

What skills and knowledge are essential for someone looking to enter the field of computer vision?

Computer vision is a vast field that requires a lot of experience to get the hang of. Understanding how the current state-of-the-art models work is crucial for someone entering this domain. Even a high-level understanding is really helpful because it gives you an idea of why things work the way they do.

Another essential piece of knowledge is how dataset preparation influences the results of computer vision algorithms. That whole “garbage in, garbage out” thing is very much true in this case.

While computer vision might call for some particular skills, the real keys to success are having the drive and passion to develop your solutions and constantly improve yourself.

Computer vision is a fascinating field. How does Addepto leverage this technology to solve real-world problems?

Addepto stays on track with the latest research, advancements, libraries, and best practices in computer vision.

This approach always allows us to find the best solution to the problem at hand. We harness cutting-edge technology to tackle practical issues across various sectors. Leveraging advanced algorithms and deep learning, we develop solutions that streamline automation and drive innovation in multiple industries.

As computer vision becomes more widespread, what industries do you believe will benefit the most from its adoption, and why?

Industrial applications have massive potential for adopting computer vision solutions to automate processes, improve quality assurance, and gain valuable insight into workflows. However, it’s crucial to implement a human-in-the-loop approach, especially in the early stages. This way, any flaws in the computer vision system can be identified, analyzed, and gradually improved until humans are only needed for supervision.

The field that could benefit from computer vision is, for example, healthcare, even though there is still valid considerable doubt regarding whether computer vision algorithms can entirely replace human experts. These algorithms lack experience, and emotions (which might impact healthcare positively or negatively), and also often exhibit bias, making it challenging to fully trust them in critical and responsible tasks.

Currently, it is safe to say that machine learning can enhance or support doctors’ diagnoses, but ongoing research will likely make the algorithms more reliable over time.

Computer vision algorithms sometimes struggle with interpreting ambiguous or noisy images. How do you handle uncertainty in the results of your computer vision models?

I always try to acquire as much domain-specific knowledge as possible. This enables me to gain valuable data insights, as data preparation and understanding are essential elements in most machine learning problems. Of course, there are methods designed to handle problematic inputs, but ensuring that the training dataset is prepared in the best way possible, covering data variations, is the number one priority.

Looking ahead, what are your predictions for the future of computer vision, and how do you envision its integration with other AI and emerging technologies?

There always has been a massive potential for computer vision, and with development in the fields like autonomous vehicles or medical image analysis, we can see that it is being utilized in the right direction. With the most recent advancements in augmented reality, it is crazy to see how well it blends virtual and real worlds.

One especially interesting idea for me is reinventing the workstation for people working with computers. Not being bound to the size and number of our monitors, but rather utilizing the full field of view of our eyes would be pretty cool! There are obviously many concerns, like fatigue, responsiveness of the system, but most importantly, security and privacy. But flying cars would be cool too.

As someone working in computer vision, how does the company create an environment that encourages experimentation and risk-taking to drive innovation?

I believe open communication and a flat hierarchy are the keys. Employees should feel comfortable sharing their innovative thoughts without fear of judgment or criticism. Accepting failures is also very important. The failures should be treated as a valuable part of the learning process. When employees feel that taking risks and trying new things is supported and appreciated, they become more inclined to continue exploring innovative solutions.

Thank you so much for taking the time to share the insights with us! If you’d like to connect with Michał and stay updated on his work, feel free to visit his LinkedIn profile.