From the course: Computer Vision for Data Scientists

What is computer vision?

From the course: Computer Vision for Data Scientists

Start my 1-month free trial Buy for my team

What is computer vision?

“

- [Instructor] I want to start with some motivation for this course. David Marr raised a simple, yet hard to answer question in his 1982 book "Vision", "What does it mean to see?" Pause for a moment and reflect on this quote. Marr noted that vision is more than knowing what is where by looking, it's discovering from images what is present in the world and where it is. We, humans, take our vision for granted. Something so easy and natural for us to do should be the same for machines but our visual system is incredibly complex and involves multiple layers of processing and interpretation that have evolved over millions of years. A unique feature of our vision is the ability for our brains to quickly learn and adapt to new visual stimuli. You and I can learn new visual concepts and object recognition very efficiently, even with limited examples. Our vision generalizes visual concepts across different objects, colors, shapes, and positions, even when occluded or when noise is present. Replicating this level of complexity and adaptability in machines was a significant challenge in computer vision research. Images and videos contain intricate and dynamic patterns, requiring sophisticated algorithms to recognize and interpret. There often needs to be more clarity and certainty present in visual data, making it difficult for machines to accurately identify objects and events. Visual data varies significantly between natural scenes, medical images, and robot vision domains. Developing algorithms that match or surpass human level performance has been one of the most challenging tasks in artificial intelligence but it is a first step towards creating intelligent machines that can learn and adapt quickly to new visual stimuli resembling more of a human-like learning experience.

Contents