Call for Code App Uses AI to Make Homes Safer and More Resilient

Disasters can hit with little warning, but often it's not the earthquakes or windstorms that directly harm people, it's the failure of substandard housing that causes the most devastation. The World Bank reports that by 2030 nearly 3 billion people will be at risk of losing a loved one or their homes to disasters such as these. However, a machine learning solution that emerged from Call for Code to help builders, local officials, and homeowners assess construction quality before and after storms might change all that.

Developers from Build Change, an organization dedicated to preventing housing loss caused by disasters, placed second in the 2018 Call for Code Global Challenge with their solution PD3R (Post-Disaster Rapid Response Retrofit). Their solution provided families with the ability to immediately assess whether their home could be structurally strengthened following an earthquake. The team received an award of $25,000 USD, and now The Linux Foundation will host an offshoot of the technology as an open source project.

With the support of IBM, Build Change has created a new open source artificial intelligence tool called ISAC-SIMO that extends PD3R technology to help builders, local officials, and homeowners assess the construction quality of newly built or retrofitted homes.

To the average person, a construction site can be a complex and confusing place. Even more so when trying to determine the quality and safety of that construction. What's the difference between a well-built and a poorly built wall anyway?

Technology and innovation can help scale access to the safe construction practices that communities in high disaster risk areas desperately need. The ISAC-SIMO solution - short for Intelligent Supervision Assistant for Construction and its Spanish translation, Sistema Inteligente de Monitoreo de Obra - uses custom IBM Watson visual recognition models trained with thousands of construction material images to assess the quality of building elements, such as masonry walls or rebar shapes. The applications for accessing and managing those models were also built on open source software, including Python Django, Jupyter Notebooks, and React Native.

When a user uploads an image, the app analyzes the structural integrity of the photographed material against the model to determine its compliance with safe construction guidelines and tells the user whether the use of that material is a 'go' or 'no go.' The user can perform multiple checks specific to their location, typical materials, and workflow. Workmanship issues can be identified by anyone with a phone, regardless of their technical knowledge of construction.

Given its potential, the technology was selected to enter the Call for Code deployment pipeline for incubation and further development to maximize its impact. The requirements for ISAC-SIMO were captured during a Design Thinking workshop held in Kathmandu, Nepal, with participants from IBM and Build Change. The team defined central personas, or stakeholders, and produced a set of user outcome-oriented project goals in support of contractors, inspectors, and homeowners participating in a Colombia government program that identifies, evaluates, and improves vulnerable housing.

Originally, the team intended for ISAC-SIMO to provide collaborative information sharing and workflow between these three users. However, due to the pandemic, the project shifted to provide tools that empower individuals to perform their own checks, which has resulted in a general framework for building quality assessments that is well-suited to be extended by the open source community.

As the IBM and Build Change teams collaborated to achieve short-, medium-, and long-term milestones to implement the solution, several technical challenges emerged. For example, how do you ensure that end users capture photos that are clear enough and positioned well enough for the machine learning model to perform? How do you ensure that the subject of a photo is detectable enough to make an accurate assessment? How do you ensure accurate sizing when quality comes down to measurements taken in millimeters?

To learn about how the Build Change team addressed these challenges, check out the IBM Think presentation replay.

The defining feature of ISAC-SIMO today is that it provides a generic open source framework for users and contributors to use the project models and leverage a growing catalog of construction quality check types contributed by the open source community to build on the solution. Already, the project has seen a boost from the Autodesk Foundation, which contributed pro-bono expertise to advise on the project's development.

'It was a great pleasure to help advance the ISAC-SIMO project with Build Change,' said Charlie G Zhang, Principal Software Engineer, Autodesk Construction Solutions. 'We had the opportunity to apply the latest technologies in image processing and machine learning to the problem at hand. I had a good time refreshing my knowledge of IM and ML and getting more handy in Python and Jupyter Notebook. From my participation in the project, I have broadened my skill set and knowledge, and had fun in the process, too.'

Eventually, the hope is that a robust catalog of pretrained and consumable models becomes available to users around the world and that this tool can truly become an intelligent aid for quality assurance and safe construction practices.

Now that ISAC-SIMO is a Call for Code project hosted at The Linux Foundation, we invite users and contributors to join the community to improve the project and help us all achieve the aim of better quality housing for everyone, particularly in disaster-prone areas.

Learn how you can contribute to ISAC-SIMO and further the development of this project.

How will you answer the call to build and contribute to sustainable, open source technology projects that address social and humanitarian issues? Get involved in the 2021 Call for Code Global Challenge.

Public link

Please use the above public link if you want to share this noodl on another website.