If you’re looking for a startup idea that could slow climate change, you could become an expert in assessing home energy efficiency. At least that’s what happened to the founders of Kelvin, a French startup that uses computer vision and machine learning to make it easier to check home energy efficiency.
Clémentine Lalande, Pierre Joly and Guillaume Sempé have started to look at energy efficiency audits for private households, as renovations will have a massive impact on reducing energy consumption and CO2.2 emissions. However, as in the rest of the construction industry, most companies in this sector do not use technology to improve their processes.
“In the next 30 years, 300 million homes need to be renovated in Europe,” Lalande, CEO of Kelvin, told TechCrunch. “But the construction industry is the second least digitized sector after agriculture.”
In France, the National Housing Authority (ANAH) has set itself the ambitious goal of building 200,000 renovated homes in 2024 alone. But the craftsmen simply cannot keep up, and that is damaging the climate. In general, the regulatory landscape for this type of startup in Europe is favorable.
Founded in October 2023, Kelvin is a pure software company. The company does not want to build a marketplace for service providers and, unlike Enter, another German residential energy consumption assessment startup reported on by TechCrunch, it does not want to be a customer-facing product either.
Instead, the startup has assembled a small team of engineers to develop its own AI model specialized in assessing residential energy consumption using machine learning. The company uses open data such as satellite imagery as well as its own training dataset with millions of photos and energy ratings.
“We calculate more than 12 proprietary, semi-public or open data sources that provide information about the building and its thermal performance. So we use fairly standard segmentation techniques and analyze satellite images with machine learning models to detect certain features, such as the presence of adjacent buildings, solar panels, collective ventilation systems and so on,” said Lalande.
“We also do this with data we collect ourselves. We have developed a remote inspection tool with a bot that tells the person on site what photos and videos to take,” she added. “Then we have models that count radiators in videos, detect doors, determine ceiling height and determine the type of boiler or ventilation system.”
Kelvin doesn’t want to use 3D technologies like LiDAR because it wants to develop a tool that can be deployed at scale. It lets you use regular photos and videos, meaning you don’t need a current smartphone with a LiDAR sensor to record the details of a room.
The startup’s potential customers could include construction companies, the real estate industry or even financial institutions looking to finance renovation projects – financiers in particular may value accurate assessments before making a decision.
In the company’s first tests, the accuracies of residential energy ratings compared to traditional ratings were within 5%. And as the system becomes the standard tool for these tests, it will become much easier to compare one home with another, and one renovation with another.
The startup has now raised €4.7 million ($5.1 million at today’s exchange rate), with Racine² leading the round and a non-dilutive investment from Bpifrance. Seedcamp, Raise Capital, Kima Ventures, Motier Ventures and several business angels also participated in the round.