categories

HOT TOPICS

NEWSLETTER

If you are considering becoming a 1M/1M premium member and would like to join our mailing list to receive ongoing information, please sign up here.

Subscribe to our Feed

Thought Leaders in Artificial Intelligence: Florian Eggenschwiler, Chief Product Officer of Xovis (Part 1)

Posted on Friday, May 27th 2022

A fascinating conversation about how Xovis is applying AI to modeling human behavior in public spaces.

Sramana Mitra: Let’s start by giving our audience a bit of your background as well as the background of the company.

Florian Eggenschwiler: I’m the Chief Product Officer at Xovis. Xovis is a Swiss-based company helping the world rethink people flow. We do that by deploying our own 3D sensors that understand people’s movement. Then we have a variety of software applications that make sense of this data to help a variety of industries to understand how people move through our physical world and initiate change to optimize the flow of people.

Sramana Mitra: Where are you coming from? What’s the thinking that has led you to this problem?

Florian Eggenschwiler: I started out in aviation operations. I worked for a number of ground handling companies. My last position was Head of Innovation for a large ground handling company. One of the products we used was Xovis to understand service level agreements between the ground handler and the airline. I thought it was just an amazing product. I joined Xovis in a sales-focused role. As of summer 2021, I moved to the position of Chief Product Officer.

Sramana Mitra: It sounds like there’s a particular use case through which you discovered Xovis. Let’s start there and double-click down on that. Get into more granular detail on exactly how Xovis impacts this process.

Florian Eggenschwiler: The initial problem is for an airport to understand wait times and security for a variety of reasons. One reason is the passenger experience, another is to better plan and deploy staff in an airport operation. With conventional technology, that wasn’t possible for a number of reasons. We set out to develop a purpose-built sensor for this.

With this, we were not only able to identify people and have an accurate count, but we are also able to deploy a number of algorithms to associate a person with a particular queue and then measure a variety of parameters. It can be how many people are in a queue, how long have they been queueing, and what are specific process times. This was previously done manually with stopwatches either by the video footage or by manual measurements once a month.

Sramana Mitra: Did I hear you say you are actually modeling individuals? Is there computer vision involved?

Florian Eggenschwiler: The technology is based on 3D vision, which means you have two lenses that create a 3D image of the surface area beneath. Then using an algorithm based on that depth, we are able to determine what is a person versus another object. Based on this, we are able to deploy neural networks to determine additional attributes such as the gender of a person. More recently, are they wearing a face mask? Are they by themselves? The retail sector is very much interested in different attributes of a person.

Sramana Mitra: I’m still stuck on the use case that you started with which is the airport wait time. To understand if a person is waiting in a queue for a long time, you have to identify an individual and then follow that person’s experience. Your model is capable of doing that?

Florian Eggenschwiler: Correct. We can connect a large number of sensors together. They get a bird’s eye perspective. We plan them in a way that there’s a small overlap in the surface area that they can cover. When an individual walks from the coverage area of one sensor to the next, there’s a virtual handshake where we pass this person as a completely anonymous dot. We don’t know who this person is. They just get an ID on our system.

Based on this, we can then draw zones or count lines where we can measure how long a person was in a certain zone or how long it takes to go from point A to point B. This becomes the basis to run more advanced algorithms to determine KPI such as wait time in a security line. Most often, it’s not the individual wait time of a person; it’s the average wait time in a 15-minute period.

Sramana Mitra: It’s operations research optimization.

Florian Eggenschwiler: Exactly.

This segment is part 1 in the series : Thought Leaders in Artificial Intelligence: Florian Eggenschwiler, Chief Product Officer of Xovis
1 2 3 4

Hacker News
() Comments

Featured Videos