How AI and Deep Learning Affect Construction

The construction industry is one of the oldest professions globally, but technology and the industry continue to evolve as time passes. The designing, planning, and building processes have changed to meet modern and technological standards.

AI technology has made a massive impact on the construction industry by helping with manual labor and essential tasks. Technology is being used more and more to facilitate efficiency, and artificial intelligence has created a new dimension for the industry to build and design projects.

While RedTeam doesn’t work with AI technology directly, we still like to inform construction professionals on what is happening in the industry. That’s why we talked with special guest Johnny Maghzal, product manager from Togal.AI, to talk about how the company leverages artificial intelligence (AI) tools, more specifically deep learning, for the construction industry.

How the Fastest Estimating Takeoff Software Uses AI to Build

Togal.AI solves the most time-consuming part in the bidding process––estimating takeoff––by utilizing cutting-edge deep machine learning technology to analyze blueprints. The construction AI engine uses AIA measurement standards to automatically and accurately detect, label, and measure project spaces & objects within seconds.

The company’s software helps save time and money by fully automating the quantity takeoff process, reduce human error by eliminating the laborious manual takeoff process that requires drawing polygons, lines, and dots to measure areas and count objects and expedite the bidding process by submitting bids faster to win more work.

Togal.AI was created to solve the pain points of the laborious estimating takeoff process. Most of the time is spent on drawing, coloring, cutting, and tracing spaces. On top of this, there’s also sculpting, speaking with subcontractors, having the right pricing, and many other essential tasks. Estimators don’t have to worry about spending countless hours drawing and coloring blueprints with the estimating takeoff software. Togal.AI does it automatically and with efficiency.

What are Artificial Intelligence (AI), Machine Learning, and Deep Learning?

AI, simply put, is machine conducting tasks that are usually done by humans. For example, robots, computer vision, predictive models, and chatbots are all AI. Anything that mimics human behavior is classified as AI.

You can think of AI as the object or thing doing the work, so robots, chatbots, models that show predictions, etc. But what makes them do the work? How do they know what they are doing? This is where machine learning comes in.

Machine learning is the idea where you put data in an AI product using an algorithm that humans engineered. AI then takes the algorithm and uses the information to complete tasks. We build a robot, and we tell it what to do. Simple enough.

Now comes the more complex aspect: deep learning. Deep learning uses neural networks, as many as four or five layers (sometimes more), and acts as a black box for the user. The engineer who builds the deep learning algorithm doesn’t know what’s happening inside the AI product.
The only thing that matters is the output, i.e., what the AI will do. Skynet? Just kidding. Maybe.

How Does Togal.AI Use These Three Aspects for Takeoff?

So, the AI is built (estimating takeoff software), the machine learning algorithm is made for it (object detection algorithms), and the deep learning neural networks are interpreted by software. How? We’ll never know. But at least they provide accurate object detection by identifying furniture, clothing, electricals, and much more.

Object detection algorithms can identify safety workers by what they’re wearing, such as the vest, a hard hat, a cone next to the individual. Togal.AI uses the same algorithm, but with different naming data, the AI can identify objects within a blueprint for estimating takeoffs. The software can identify the number of bathtubs, sinks, single swing doors, double swing doors, openings, and more to provide an accurate estimate.

Image Segmentation’s Role in Identifying Objects

In segmentation, the algorithms work at the pixel level. So instead of predicting what a collection of pixels are representing, they try to predict the category of every single pixel in the provided space.

There are two different types of segmentation, semantic and instance. To explain the two, we’ll talk about cats and dogs. Imagine an image with one dog and three cats sitting side by side. Semantic segmentation will identify the dog and cats by tracing boundaries around them (tracing their bodies). It identifies which animal is the dog and the other animals as cats.

For instance segmentation, it takes identification to the next level by identifying which animal it is and also counting them by classifying them as “cat one,” “dog one,” “cat two,” and “cat three.” The segmentation outlined the animals and counted them at the same time. This is the type of technology Togal.AI uses when reading plans and recognizing objects. The software will know five sinks or seven single doors and render the data to produce accurate estimating.


To learn more about how you can incorporate RedTeam into your business development – project management – project financials, click here to watch and learn more about our 5 MINUTE VIDEO OVERVIEW!