Word Count: 1,712 Estimated Read Time: 7 Min.

Artificial Intelligence and Image Recognition

Despite the avalanche of platforms and programs using Artificial Intelligence (AI) that have debuted in the last year, AI is not really “new.”  In fact, it has been a long time coming.   As early as 1949, Donald Hebb had developed Hebbian Learning, a possible algorithm for learning in neural networks, which is what today’s AI programs use.  And by 1955, Arthur Samuel at IBM had written the first game-playing program for checkers that could “learn” to play.  A computer that can learn is, by definition, artificial intelligence.  But those innovations were preceded by a century of mathematical and computational advances.  And, it took another 75 years for mathematicians, programmers, engineers and linguists to advance the field of computing and neural networks to produce the advanced AI platforms emerging now.  So for anyone paying attention, the advent of AI was surely no surprise.

And yet, now that Artificial Intelligence is here, the world does not appear to be truly ready for how it will change the world creates, works, learns, plays and innovates.  No doubt, it will revolutionize all industries and areas of society.  AI, coupled with advances in robotics, will revolutionize how things are done.  But today, for the average person, a change this monumental can be hard to grasp.  How do business leaders wrap their minds around the seismic shift on the horizon?

Of course, there is a tsunami of books, articles, posts, podcasts, and programs focused on explaining what AI is and what AI platforms can do. But it is a lot to take in.  Given this tidal wave of information, how does a busy business owner capitalize on the benefits of AI now?  Perhaps the best way is to break the information down into major tasks that AI can do better and faster than humans and see what benefits might apply to his organization.  Let’s start with image recognition.

AI Image Recognition

Image recognition – which is part of what is considered ‘computer vision’ – is the branch of AI that recognizes, analyses and interprets photographs to identify objects, places, people, or things.  The major goal of AI image recognition is to view the objects in the same way that a human brain would.  It detects and evaluates a myriad of details — examining each pixel in an image to extract relevant information in the same way that humans do — and then draws conclusions based on that analysis.  AI cameras can detect and recognize a wide range of objects trained by computer vision.  (Computer vision is the broader field that encompasses the ways that data from the world is acquired, analyzed, and processed and then fed to machines.)  So image recognition is a lot like our eyes ‘seeing’ the world and computer vision is similar to our brains processing what our eyes see. 

Over time, the more images the algorithm recognizes, the better it is able to predict (with increased accuracy) what is in the picture.  That is the process of training the algorithm.  For instance, the algorithm might look at the photo of a dog.  Initially, the image recognition system might indicate that — because of similar characteristics — there is an animal in the image and it is 85% likely to be a dog, 10% likely to be a cat, and 5% likely to be some other similar animal such as a wolf or coyote.  This is referred to as a confidence score.  In order to accurately anticipate the object, the machine must first grasp what it sees, then analyze it by comparing it to past training to create the final prediction.  But the more images of dogs the machine sees as well as images of cats, wolfs and coyotes, the better able it is to distinguish between them.  This is not unlike how young children learn to distinguish between animals.

Image recognition can be used in a variety of applications, such as labeling the content of images with meta tags, performing image content search, and guiding autonomous robots.  It can also be used by self-driving cars and accident-avoidance systems.  And it can be used for facial recognition in medical diagnosis and by law enforcement to identify criminal suspects in video recordings or missing persons.  There are so many ways that businesses can use image recognition to its benefit.

Using Image Recognition for Business

To better understand how image recognition could be useful in business, let’s explore how image recognition is already being used by the real estate industry to benefit both buyers and sellers.

1. Property search and valuation – One use of image recognition in real estate is to improve the property search and valuation process.  Buyers can use image recognition to filter and rank properties based on visual features, such as one-story home vs. two-story home.  While sellers can use image recognition to estimate the market value of their properties based on the comparison with similar listings, such as homes with pools and screened patios.  Image recognition can also help detect and highlight defects or damages that might affect the value of a property, such as fire damage in the interior of a home or roof damage after a hail storm.  It can streamline processes, improve user experiences, and provide valuable insights.
   
   
2. Virtual Tours and Virtual Staging – Image recognition is also being used to create and enhance virtual tours. Virtual tours offer buyers immersive experiences, allowing them to explore a property online without physically visiting it.  Image recognition helps create realistic and detailed virtual tours by stitching together multiple images and adding depth and perspective.  The seller can create a photorealistic 3d virtual model — enabling 360-degree views and captivating walkthroughs — that takes property marketing to the next level.
   
   And with virtual staging, the seller can add or remove furniture, decorations and accessories to a property image to make it more attractive and appealing.  Image recognition also helps create virtual staging by recognizing the style and theme of a property and suggests suitable items and placements.  This is particularly helpful for properties that are vacant by showing the placement of furniture in rooms that are empty so prospective buyers can visualize creative ways for a home’s space to be utilized.
   
   
3. Construction Quality – Image recognition is also being used to analyze the quality of construction.  Realtors and inspectors are using well-trained image classifiers to check if the construction quality is up to the mark.  It is also being used to spot the weak structural areas which can pose a threat to the structure.  This can be particularly helpful, for example, in areas impacted by an earthquake or flood.  A well-trained image classifier can catch small issues that even a trained professional might miss or overlook. 
   
   Case in point.   Flooding in the parking garage, combined with the salty sea air, had created a unique set of structural challenges at Champlain Towers, a 12-story beachfront condominium in Surfside, Florida.  There were visible signs of decay but city building inspectors did not conclude that structural failure was imminent.  Neither did the Condo Board or the building’s residents.  While they all thought it was bad, no one thought it was so bad that the building was about to collapse.  Before repairs could be done, the building collapsed in 2021 resulting in the tragic death of 98 residents while they slept. 
   
   It is now believed that a well-trained image classification system could have done a better job of warning inspectors, city engineers, and the building tenants at Champlain Towers about the seriousness of the building’s decay.  Had the building’s security cameras located in the parking garage been fed into a well-trained image classifying system, it would likely have signaled a critical warning.  Because machine learning uses thousands of images of building decay to learn how to identify levels of damage ranging from superficial to imminent structural failure, it would have had a “model” and could look for a multitude of minor differences that separate building decay that is superficial from decay that is structurally compromising.   The program’s “training data” would have been able to detect things that humans miss.  (Indeed, this same type of image recognition detection is being used in the medical field to examine CAT scans and detect cancer.) 
   
   
4. Security and Access Control – Image recognition is also being used in real estate to improve the security of and access to properties. Using facial recognition, biometric scans or license plate recognition, it is being used to verify the identity of visitors and tenants. This prevents unauthorized access, theft, or vandalism, and provides occupants with peace-of-mind.   In a shared or rental housing scenario, image recognition can automate the check-in/check-out process, making it hassle-free and secure. In commercial setups, it can monitor and manage the use of shared spaces such as meeting rooms or amenities, ensuring optimal utilization.  And, when it comes to security, AI doesn’t sleep, slack, or even blink.  This ensures properties are constantly watched.  And it is quick to notify property owners or managers of anomalies in activity, be it from a fire, flood, or intrusion, and ensure swift action.
   
   
5. Maintenance and Inspection – Image recognition is also being used in real estate to facilitate the maintenance and inspection of properties.  It’s used to automate and streamline the maintenance and inspection tasks by detecting and reporting problems that require attention or repair, such as cracks, leaks, mold, pests, and other defects that affect property quality and safety.  It also tracks and documents the progress and results of the maintenance and inspection activities.

This shows that just one aspect of artificial intelligence, applied to one single industry, is reaping enormous safety, security, marketing, sales and financial benefits.  And that is just what it is being used to do today.  In the future, there will be other applications for the image recognition aspect of AI to the world of construction and real estate.  While AI is a huge elephant to eat, the best way to begin is one bite at a time.  As you learn about the multitude of ways AI can make your business run more smoothly, pick just one and try it.  Once that happens, you’ll feel comfortable trying another and another to streamline your business. 

Next week, we’ll examine other aspects of AI that can benefit your business now.  Stay tuned!

Quote of the Week

“It is difficult to think of a major industry that AI will not transform. This includes healthcare, education, transportation, retail, communications, and agriculture. There are surprisingly clear paths for AI to make a big difference in all of these industries.” Andrew Ng

© 2023, Keren Peters-Atkinson. All rights reserved.