Until recently, humans have had many traits that separate us from machines, including our extremely complex sense of sight. The combination of brain and eye working together to understand what’s in front of us and provide us with the context needed to make informed decisions is extremely complex. And yet, it is already available through our phones and other devices.
Features such as speech recognition and voice ordering have become standard, but computer vision has lagged due to its complexities. Computer vision aims to give a machine or computer the ability to visually sense the world around it. This means computers are able to recognize, understand, and categorize objects. As computer vision’s impact grows, it is important to understand how it can be used and common issues that arise from using it.
Before diving into how computer vision is used, it is important to understand where the concept comes from. While rapid developments have been made in the past 10 years, computer vision has been a topic of focus since the 1960s. In 1966 Marvin Minsky, one of the co-founders of the Artificial Intelligence Lab at MIT, gave an assignment where students had to link a camera to a computer and have the computer describe what it saw. In the late 1980s and early in the 1990s, the use of computer vision for facial recognition began through the Eigenface approach. It’s taken decades for computer vision to integrate into everyday society, but we are beginning to see computer vision in common applications such as social media.
You Are Probably Already Using Computer Vision
Social media has been implementing computer vision for years as Facebook and Snapchat continue to use the camera to improve user experience. A great example of this is Snapchat’s facial overlay feature. The concept is simple. When taking a picture using the front facing camera, users hold the camera to their face allowing it to be scanned. Snapchat then processes this information and allows users to choose from a variety of filters to overlay on their face. Some favorites include dog ears, flower crown, and face-swap. Another example of computer vision at work is Pinterest’s Lens feature. Lens allows Pinterest users to take a picture of an object with their phone and then provides a list of words that describe the object as well as suggestions for similar items. You may have also seen Amazon’s camera-driven product search, or similar efforts by Google (amongst others). The concept of a camera identifying objects defines computer vision, but understanding the context of those objects is where computer vision runs into issues.
We Need More Context
Computer vision allows machines to identify objects, but the context of those objects, and what they mean in relation to other things tends to be lacking. Take for example, a farm far out on the horizon. There is a black and white dot next to a red barn. While a human could identify that black and white dot as a cow, a computer may not. The reason the human is able to identify the dot is that we understand the context of the farm. A computer on the other hand would only recognize the focal point-the barn, and would not be able to make the assumption that the dot would be a cow. This is not the only challenge that comes with computer vision, with things such as discrimination, commonality, rotation, occlusion, and lighting also causing issues.
Take for example the image above of a Lynx in a tree. While computer vision technology can recognize that the animal is a member of the feline family, it may be unable to discriminate and establish that it is in fact a Lynx and not a Cat. These issues arise as different objects can look similar, with details necessary to establish exactly what the object is. This can lead to an image tricking computers and leading to false or incorrect results.
A Vision of the Future
Computer vision is by all accounts a game-changing technology, with future applications including improved medical diagnosis, safety, and security. As computers become better equipped to identify objects, the same can be done to identify cancer and other diseases. Another example of computer visions’ future applications is the newer models of smart home devices such as Amazon’s Echo Show. These devices now include cameras, which could be used to detect if a person is injured or if a burglar has entered a home. Computer vision could also be used to read body language and facial expressions. For example, Alexa could read a person’s facial expression to be sad or upset, it could use the information to change the lighting of their room accordingly. Another way computer vision could be applied is to support disabled individuals with limited sight. A device could use computer vision to relay information on the objects surrounding a disabled person, communicating by voice, vibrations, or sounds. These applications of computer vision could be revolutionary and are made possible by their connection with Machine Learning.
Machine Learning: Computers Making Informed Decisions
Machine Learning is an application of artificial intelligence (AI) that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. Machine learning ties into computer vision when the conversation of context arises. As previously mentioned, computer vision falters at times due to a lack of context. Machine learning takes an image and information gathered by computer vision and allows a machine to determine what it should do with the information. An example of machine learning and computer vision at work is self-driving cars. Self-driving cars use computer vision to identify and detect lane lines, objects, and traffic signs. Machine learning takes this information and instructs the vehicle to respond according to the situation. Features such as, collision avoidance, and self parking are common functions of computer vision in modern vehicles. When combined with machine learning, the applications of computer vision are potentially limitless.
Ready For Your Computer Vision Project?
While the use of computer vision is on the rise, it should only be implemented when it can add value to a situation. It’s big and complex and something that requires utmost care and consideration. When considering computer vision for future projects it is important to consider the following factors.
- Does my business involve identifying or categorizing objects? This is the most natural case, as computer vision would provide information about the object instantly, removing the research an individual would do to reach the same point.
- Does my business offer products or services? Computer vision could effectively be used with either product or services businesses but how it’s used could be radically different. A product company such as a tire retailer can use computer vision technology in its app to allow customers to identify the size and trim of their current tires. On the other hand, a services app that identifies and provides information on wild plants could be used by campers looking for non-poisonous food.
- Does computer vision contribute to eliminating steps for the target audience? For example, a grocery store could use computer vision to identify products that users want to purchase online or in their next trip to the store. How it would work is that a customer would use their grocery mobile app to take a picture of a product they want to purchase, such as cereal. The app would take a picture of the cereal box and identify that the box is Kellogg’s Frosted Flakes. From here computer vision would work with databases and machine learning to provide users with up to date pricing on the product. Customers can then choose to add the object to their online ordering cart for their next pickup.
Computer vision is only one piece of the evolving technology impacting everyday society. Pretty soon we’ll have robots doing more than just backflips, and computer vision will be a massive reason why.