Case study: Fighting fire with modern technology

OVERVIEW
Wildfires in the western United States ravage an average of 1.5 million acres each year, the rate and size of blazes increasing as climate change feeds the warmer and drier conditions the burns require – with little relief in sight. A recent study by a team of researchers estimates that some 3.3 million U.S. households — equal to over $11 billion worth of property — will be at risk of wildfire damage each year over the next three decades. Pyrezo – a start-up tech company in Bainbridge Island, WA., –  developed a way to leverage simple image capture and computer vision to help property owners, HOAs, and fire protection districts map, track and reduce wildfire risk. In a relatively short time, Pyrezo had established the technology needed to collect and analyze intelligence to mitigate wildfire risk with a desktop application. To appeal to more users and streamline the image captures needed by the AI-aided program, the next step was developing a cloud-based, mobile app solution.

WHAT WAS NEEDED

Here's how Pyrezo works: Users upload a series of images captured by a camera from several angles around a property. Once inputted, those images are analyzed by both humans and an artificial intelligence filter to identify vegetation and assess the environment on and around structures, looking for conditions that might increase the risk of wildfire loss. The data formulated from that analysis is compiled into a "Wildlife Risk Reduction Report" – essentially a prescription landowners can follow to minimize their risk of fire.
"For someone who lives on the East Coast, I wasn't familiar with wildfires," said Hilel Tcherikover, a Senior Technical Project Manager for ASMBL. "I never imagined there would be a platform that could help mitigate risk."
That platform already existed in Pyrezo’s desktop version, but it required an external program and additional tooling to complete the important photo annotation tasks. It did mean ASMBL wasn't starting from scratch in moving the program from a desktop to an iOS app, and it made clear the need to accelerate and simplify the annotation process.
"The groundwork that went into this was very impressive," Tcherikover said. "The tech process was kind of laid out. We had to get a desktop application into the cloud. We had to ensure communication from the platform's front end to the back end and tie that all together in a lens of efficiency."

Yes, efficiency: As a startup, Pyrezo had tighter timelines and budgets than some of the large tech companies that have enlisted ASMBL's help in the past.

"We started the project in January and the goal was to complete it by the end of March," said Tcherikover. "We had to work very quickly. Part of the reason for that timeline constraint is they wanted to launch in time for fire season."

THE APPROACH
Converting the desktop version of Pyrezo to a cloud-based app came with unique technical needs, said Tcherikover.

"We had to take a desktop product and put it into something that looks and acts dramatically different from what it did on the desktop," Tcherikover said. "All of the information that was stored within the desktop application had to be accessible via the cloud and we had to streamline the annotations on the photo that had to be completed. "
In creating the iOS version, the ASMBL team eliminated the need for manually annotating photos through an external process. They also adhered to a five-second processing rule, meaning the app’s programming allowed each photo to be scanned and inputted in five seconds or less.
The need to get the app functional in a short period also necessitated the ASMBL team to know what to leave in and what to leave out.

During the development period, said Tcherikover, ideas about additional features and functions bubbled to the surface. But, in the interest of time, rather than tackle them as they cropped up, the ASMBL team set them aside with plans to fold them into an updated version of the app.

THE SOLUTION

Available in the Apple App store, Pyrezo Capture allows users to easily take and upload photos of their property. The on-phone guide provides easy-to-follow instructions for taking photos from several vantage points around the property including vents, windows, and roofs. Once the images are uploaded and annotated, a risk report is generated.
One of the crowning achievements of the ASMBL team’s work was figuring out a workaround for the external hardware central to the photo annotation process in the desktop configuration. Getting that onto a website, said Tcherikover, meant simplifying it into a small combination of keyboard commands and keyboard clicks. Even being in the cloud, the app version receives and inputs each photo in less than five seconds.

The app is intuitive. The uploading of images to be analyzed by AI is an easy process. To date, there haven't been any major bugs.
"I've rarely worked with tech that doesn't come back with at least one bug, we went through the quality assurance phase without any," Tcherikover said. "I don't know if I've had that experience. It was kind of shock, how is there not one issue?"

While the 2023 fire season is behind us, Tcherikover said more users are downloading the app because there's no bad time for homeowners to lessen their risks. "I've seen the number of downloads grow week after week," said Tcherikover. "From all indications, the app is doing exactly what it's supposed to do.