The world’s first end-to-end
physical AI for nature

Deep Forestry is a physical-AI company in nature. Our robots autonomously capture ground truth where no satellite or person can: beneath canopies, along corridors, across cluttered and GPS-denied ground. Our own AI turns it into measured, traceable objects. Customers use it for inventory and yield, carbon and MRV, vegetation and asset risk, and mapping and navigation. Deep Forestry AB is based in Uppsala, Sweden.

No. Nothing in our stack is specific to forests; forests are just the hardest place we've proved it. Today we work on forests, plantations (palm, eucalyptus), power-line corridors and contested, GPS-denied ground. Mines, cities and infrastructure are next and not yet deployed.

Organisations whose decisions depend on ground no one can see well from above:

• Land, forest and plantation owners and managers who need a measured inventory
• Carbon project developers and verifiers who need numbers an auditor can trace
• Utilities managing vegetation along power lines and other critical infrastructure
• Satellite, earth-observation and Geo AI companies that need ground truth to calibrate their models
• Organisations that must map and move across GPS-denied ground

Neuron Edge™ moves a drone or backpack close to the ground, among the things that matter, scanning with LiDAR as it goes. Neuron Mind™ then finds and separates each object in the point cloud and measures it; in a forest or plantation, that means every tree's position, height, stem diameter and volume. Because we capture from inside the terrain rather than from above, we measure what top-down sensors can only estimate.

Located, measured objects, plus a terrain model and a georeferenced point cloud, as files your systems can read and in an application your team can use. What's measured depends on the output you need:

• Forests and plantations: every tree the survey sees, with position, height, diameter along the stem (measured where seen, fitted above), stem volume, density and basal area
• Corridors: each tree's height, lean and distance to the line
• Mapping: a georeferenced 3D map and terrain model of ground maps from above can't show

Accuracy depends on the terrain and what's being measured, so rather than quote one number for everything, we report how well each object was seen and keep measured and estimated values apart. Our deepest validation is in forests, where measurements have been compared tree by tree against harvester measurements in radiata pine. The best test is your own: survey a site where you have reference data and compare.

For many decisions, top-down data is the right tool: it covers huge areas cheaply. But it sees the top and estimates what's beneath. When a decision needs what's underneath, such as volume, carbon or the exact position of an object, it needs ground truth. We don't replace satellite and airborne data; we're the layer beneath that calibrates and verifies it.