How satellites create enterprise opportunity for geospatial machine learning

As the cost to launch satellites and drones has fallen recently, satellite and aerial imagery has become far more affordable and accessible. Switzerland-based Picterra, whose no-code machine learning platform allows enterprises to gain insights from earth observation imagery, is looking to seize on the opportunity to help companies anticipate and mitigate risk on a global scale.

The company announced today a $6.5 million raised from its series A funding round — part of a growing global aerial imagery landscape that is projected to be nearly $5 billion by 2026. 

“Today, the total of all satellites collect between 200-300 terabytes of imagery daily – compared to something like Wikipedia, which, in all of its languages, is only 80 terabytes of data,” said Pierrick Poulenas, CEO and cofounder of Picterra. “The only viable way to extract insight from this huge amount of data is to use machine learning and AI.” 

The geospatial machine learning landscape is currently dominated by consultants who manually train algorithms, he said, adding that most of these projects fail. “Say you want to train an algorithm that will be able to count trees [on] the surface of the planet,” he said. “First you need to create a huge training data set made of tree pictures to train the algorithm, then a machine learning model from scratch to input the training data.” But this method can be inefficient, taking months to complete. And once the model is deployed on an enterprise’s IT infrastructure, it can also introduce deviation and bias, leading to insights that aren’t useful. 

Other startups use machine learning to automatically train algorithms in this market, but Poulenas claimed that Picterra is the only one to develop the workflow into a single product. According to the company, its no code machine-learning software-as-a-service (SaaS) platform allows both technical and non-technical users to “train, manage and deploy powerful geospatial algorithms that rapidly transform images into real-world positive impact.”

ESG sector is important use case

One of Picterra’s important use cases is in the ESG sector, as companies look to prove their reporting claims as well as anticipate and mitigate risks associated with climate change. Their global customers, which include SGS, CYIENT, Westwood and The World Bank, proactively monitor, among other things, transport, infrastructure and energy networks. 

Earth observation imagery has also always dealt with land observation, mapping and management. Another Picterra customer, Nespresso, monitors coffee plantations to make sure its 1,000 farms grow coffee in a sustainable way and, as part of its commitment to building sustainable farming communities, that farmers don’t only rely on coffee to make a living. 

“Gaining insights from earth observation imagery only makes sense on a large scale,” said Poulenas. “Nespresso wants to be able to report on the farming practices of those 1000 farms in a consistent manner.” 

As investors look to take advantage of the ESG reporting trend, there is a growing alignment between what the geospatial machine-learning technology can do, the needs of the market and the funding that is getting into it. For example, “tracking deforestation has been something we have been working on since the early days of Picterra, but back then tracking, say, illegal logging activities in West Africa was seen more through the angle of taxation,” he said.

“Now the perception is different — customers are really tracking deforestation for its impact on biodiversity and so on, while they also find financial value” when it comes to ESG reporting pressures and potential new requirements — such as the SEC’s proposed rules to standardize climate-related disclosures. 

Anticipating risk at a global scale

Picterra is also finding financial value in tracking supply chain issues by analyzing earth observation imagery. “With the constraints on the supply chain, as we have seen during the pandemic or due to climate change, enterprises can get a snapshot of what is going on at a global scale, such as knowing where their containers are around the world,” Poulenas said. “Global companies that source raw materials and convert them into consumer goods need to be in control of everything in the supply chain.” 

Overall, the biggest enterprise opportunity “is the ability to use earth observation imagery, combined with machine learning, in an effective way to anticipate risk and mitigate risk at a global scale,” he said.