Humans have five senses that allow us to consciously experience the natural world. Computers are evolving at an unprecedented pace, adapting senses that took millions of years to evolve naturally in humans. From once considered lifeless machines, computers are developing vision, which bridges the gap between what is a machine and what isn’t, and inches ever closer to the singularity. Computer vision is an algorithm that has the ability to observe digital imagery or videos, and extract a description, and categorize objects. Computer vision arose as a possibility in the 1960s, where primitive versions of neural networks deciphered squares from circles.
Advancements In Computer Vision
In the 1970s, computer vision adapted to be able to read the typed or handwritten text. Only in recent decades has it been the subject of rigorous research. Deep learning technology has allowed advances in computer vision, and by 2022, the computer vision market is expected to exceed $48.6 billion.
Wayne Thompson, Data Scientist from SAS Institute, said “Computer vision is one of the most remarkable things to come out of the deep learning and artificial intelligence world. The advancements that deep learning has contributed to the computer vision field have really set this field apart”. The recent trend in computer vision is attempting to emulate the human visual system. The camera is the eye equivalent, and the computer algorithm corresponds to the visual cortex of the brain.
Main Challenges Computer Vision Has To Face
The challenge is that we don’t understand how human vision works. Biological organisms are becoming the frontier of what technology tries to replicate since organisms are the most successful, efficient, and complex algorithms in the observable world. However, our understanding of biology is somewhat primitive, as the field is shrouded in elusiveness. We cannot understand consciousness, therefore replicating digital consciousness is near impossible without concrete scientific understanding. Artificial Intelligence expert Jason Brownlee said “Perceptual psychologists have spent decades trying to understand how the visual system works and, even though they can devise optical illusions to tease apart some of its principles, a complete solution to this puzzle remains elusive”, therefore laying the challenge for computer vision. The overarching architecture of the visual pathway is extremely complex, perplexing scientists, and remains a mystery.
Giant Amount Of Visual Data
Another challenge is the sheer amount of visual data in the visual field, and how objects can be interpreted from different angles of perspective, texture, focus, different lighting, and contours. Our brains have developed semantic networks for categorization, but teaching a computer how to visualize and determine different objects is not black and white.
How Does Computer Vision Work?
Computer vision relies on convolutional neural networks (CNNs) which are mathematical, and convoluted as the name suggests. They are classes of deep neural networks, connected in a way similar to that human neurons are interconnected. Furthermore, they are encoded based on a red-blue-green encoding, in order to perceive similar to humans. They are trained using thousands of sample images, identifying pixel by pixel patterns, and memorizing the ideal output of the image, corresponding the picture with the outcome.
Each Picture Refines The Model
Computer vision operates at a pixel level, each pixel is assigned a weighting depending on its color and brightness. Each picture iteration the algorithm is exposed to further refines the model of predicting the outcome of the image. Therefore, if you wanted an AI to learn how to identify a cat, you would input millions of cat pictures into the algorithm, and it would create its own identification model, characteristics that may seem unorthodox. Attempting to program whiskers, a tail, pointy ears into an algorithm will cause failure to differentiate between a cat, and a human in a cat costume, since you would need to account for all of the nuances that the computer vision algorithm notices.
Trial And Error Algorithm Learning
Deep learning is used to refine. If we send million images of different flowers, the algorithm learns via trial and error, to refine its categorisation of what is a flower and what isn’t, creating a model. Therefore, after a while, it would be able to predict based on its ever refining model of ‘flower’. This is the foundation of a neural network, extracting patterns from data samples, which is similar to how a human learns. It can therefore predict with a probabilistic confidence level. Some computer vision algorithms have a 99% accuracy, accounting for machine learning.
Applications Of Computer Vision
Applications of computer vision are vast, extending past the mere facial recognition technology of unlocking your iPhone. Autonomous vehicles use computer vision to make sense of their surroundings, using video capture in real-time to analyze and categorize pedestrians, traffic signals, vehicles, and other objects.
Autonomous Vehicles
Over 1.25 billion traffic collision deaths occur yearly around the world, and the US department of transportation outlined that 94% of fatal crashes are attributed to human error. This outlines the importance of autonomous vehicles, using computer vision and AI to follow traffic rules.
Healthcare Sector
Healthcare sector, computer vision is being used in automated tasks for cancerous tumor detection from skin images in MRI and CT scans. They can identify small nuanced patterns unrecognizable to humans, with a 95% accuracy, exceeding trained radiologists with a 65% accuracy. Engineer professor Ulas Bagci said that “Lung cancer is the number, one cancer killer, in the United States and if detected in late stages, the survival rate is only 17 percent. By finding ways to help identify earlier, I think we can help increase survival rates.”
Computer vision reach is extending across many different areas, which is expected to continue into the foreseeable future as technology slowly integrates itself into our lives. Computer vision lays the foundation for replicating organic organisms that nature has crafted. We are essentially leading towards playing the hand of God, in creating a vision, one of the fundamental senses that has allowed us humans and other animals to experience life.
