Deep Learning is a Subfield of ML and is Concerned with Algorithms

In a Conversation with Jyoti Bhagat, Assistant Editor, TimesTech Buzz, Mr. Paul Wallett, Regional Director for Trimble Middle-East and India region speaks on how Deep Learning and AL/ML is taking the Construction sector to another level.


TimesTech Buzz: What is deep learning and how does it relate to machine learning/ AI?

Paul: All Machine Learning (ML) at its most basic is the practice of using algorithms to analyse vast volumes of data and learn from it, and then make a determination or prediction about something. In other words, rather than hard-coding software routines with a specific set of instructions to accomplish a particular task, the machine is “trained” and made to learn how to perform the task by using large amounts of data and algorithms.

Deep learning is a specific type of machine learning that focuses on a specific type of machine learning algorithms, called a neural network. What makes it “deep learning” is that there are many ‘layers’ in the neural network that the user can pass some sort of input through, such as an image, audio clip, or bit of text. Modelled on the human brain’s neuron network, Deep Learning is thus a subfield of machine learning and is concerned with algorithms inspired by the structure and function of the brain, which are called artificial neural network.

ML can thus be termed as an approach to achieve artificial intelligence, and Deep learning is one of the techniques to implement machine learning.

TimesTech Buzz: What makes deep learning applications special as compared to other types of machine learning?

Paul: A big advantage with deep learning is that it is powered by massive amounts of data. This fact is also a key part in understanding why deep learning is becoming popular. The “Big Data Era” of technology that we are stepping into, with billions of IOT devices each generating high volumes of data, provides unprecedented opportunities for new innovations in deep learning.
Further, the deep learning technique models itself on a human brain and thus behaves like how humans learn things. To illustrate with the example of how we learn a new language, an artificial neural network learns categories incrementally through its hidden layer architecture, defining low-level categories like letters first then little higher level categories like words and then higher level categories like sentences.

The magic of deep learning is that even as the model gets bigger, even as you’re passing through more and more layers — asking more and more questions — the accuracy is not decreasing. That’s not normally true for any other machine learning algorithm. To take another simple example, a construction modelling system like BIM can use deep learning to interpret less than perfect drawings and still render them professionally, based on relatively limited new input and a huge backlog of previous data.

TimesTech Buzz: What are some practical applications of deep learning in construction?

Paul: We are already seeing the first wave of AI and ML applications disrupting the way construction workflows are executed onsite. Deep Learning is now poised to the next big thing and it will further coincide with the construction industry evolving in sync with the Industry 4.0 paradigm.

One of the most widely recognised applications of Deep Learning in the AEC industry is Automated construction site equipment monitoring. We already have the technology available to track equipment and vehicles. But, for it to work everywhere, all the time, everyone and everything on the site would have to be tied to one system with an RFID tag or GPS beacon on every piece of equipment, which is easier said than done.

With a deep learning algorithm trained to understand what an excavator is or what a crane looks like, it could look at a real-time video feed and show you the live location of the excavator. If you add vehicle number plates or any other visual identifiers, the algorithm can track a specific excavator or crane throughout the day – telling you how productive your equipment is on a site. And if a site has multiple cameras to achieve depth perception and spatial relationships, we can use deep learning on a construction site to say exactly where things are in the real world.

Perhaps one of the most important application is As-built site data: One of the most powerful benefits of BIM-powered field technology is the ability to bring digital information into the built environment in numerous ways. And, data from the built environment can be digitized and manipulated just as easily. Deep learning can be combined with total stations, cameras, drones, or robots to completely automate onsite processes like scanning and quality inspections which increases the accuracy of As-built site Data.

TimesTech Buzz: Can a deep learning machine design an entire building?

Paul: Yes. If a deep learning artificial neural network has analysed and imbibed enough number of buildings and knows all of its information, then it might be able to work backward and design an entirely new building on the basis of analysed data and pre-defined preferences.

Building design optimisation is in fact a key application of Deep learning, and the architects and engineers stand to gain tremendously from the unique data-crunching abilities of an artificial neural network.

For example, a building modeling system could interpret less-than-perfect drawings — even simple pen doodles — and render them professionally based on its understanding of what you meant through analysing a huge backlog of previous data. An architect could draw a rectangle in a certain position and the system would magically fill in a complete window, or it would bring up a list of 10 different windows they have designed in the past, allowing them just select which one they want. It’s so much quicker and simpler than digging through past content files and searching for a window design made by so-and-so and of this size.

Taking it a step further, the building modeling software could take those 2D points and instead of building a 3D model, generate content ready for a 3D Augmented Reality (AR) viewer right out on the construction site.

TimesTech Buzz: How ready is Indian construction industry to embrace Deep Learning tech in their workflows?

Paul: According to Zion Market Research, the global AI-in-construction market is expected to reach $3.1 billion by 2024, with a CAGR of 38.14% between 2018 and 2024. While the events of 2020 may result in some amendments to that forecast, the ability of AI and ML to support evolving workflows and address labor shortages can’t be ignored, especially for forward-thinking construction companies needing to navigate a potentially long road to recovery.

As widely reported in Indian media in 2019, Larsen & Toubro announced a new strategic initiative, named L&T-Nxt, to focus on Artificial Intelligence (AI) virtual reality (VR), augmented reality (AR) amongst other cutting-edge technologies, to drive the change in industry’s adoption of construction tech.

Other leading construction companies in India have also shown great keenness in exploring AI, ML and deep learning technologies to create new competitive advantages for themselves in India and globally. Many of these companies compete with global construction players in bidding for prestigious international projects and therefore realize the value of using cutting-edge technologies to win these projects and execute them successfully.

As we speak, there are a few pilot projects underway in different parts of the country that are actively deploying these technologies at actual construction sites. We expect this trend to gain momentum in the coming months and quarters; particularly in the wake of COVID-19 pandemic as construction industry prepares itself for the next normal.