Learning path
Back to basics: the maths and science behind GeoAI, 3D and VR
AI can draft the code, the model and the scene. People are still needed to develop, fine-tune and put them in context, and that takes the fundamentals.
- Length
- 10 articles · about 66 min
- Level
- Beginner to intermediate
- Before you start
- School-level maths. Python helps for the hands-on tutorial.
Best for: Anyone who wants to understand, check and adapt what AI builds, not just run it
By the end
What you will be able to do
- Use vectors, rotations and coordinate frames with confidence
- Explain datums, heights and why degrees are not flat distances
- Reason about light, spectra and why a model confuses shadow with water
- Judge a map or model with accuracy, recall and honest testing
- Understand photogrammetry, human perception and comfort in VR
- Review AI-generated GIS, 3D and ML code before you trust it
The path
Read in this order
Concept · Beginner · 6 min read
Back to basics: why maths and science still matter when AI can write the code
AI can build a map, a model or a 3D scene in minutes. Knowing whether it is right for a real place still takes a person who understands the fundamentals.
You will learn
- AI is excellent at producing a first draft. It cannot tell you whether the draft is right for your place, your sensor or your users.
- People are needed to specify the problem, check the result, tune it and put it in context.
Start here →
Concept · Beginner · 7 min read
Vectors, coordinates and transformations: the maths behind every map and 3D scene
Points, directions, rotations and scale: the small set of ideas that sits under GIS, game engines, VR headsets and AR.
You will learn
- Every location, pixel and 3D vertex is a list of numbers, and a vector adds direction and length to it.
- The dot product measures alignment, and the cross product gives a perpendicular direction.
Read →
Concept · Intermediate · 6 min read
The Earth is not a sphere: datums, heights and distance for GeoAI
Why coordinates need a datum, why GPS height is not sea level, and how to measure distances on a curved planet.
You will learn
- The Earth is a slightly flattened, lumpy ball. Maps use a reference ellipsoid and a datum to describe it.
- The same coordinates in two datums can differ by tens or hundreds of metres.
Read →
Concept · Beginner · 6 min read
Light, spectra and the physics behind remote sensing
Why vegetation looks bright in near-infrared, why shadows look like water and why a model trained on one image can fail on another.
You will learn
- Satellites do not see objects. They measure light, and that light has travelled through the Sun, the atmosphere and the surface.
- Different materials reflect and absorb different wavelengths, giving each a spectral signature.
Read →
Concept · Intermediate · 6 min read
Statistics and uncertainty: how to tell whether a map or model is right
Accuracy, precision, recall and why a model that is 95 percent accurate can still be useless.
You will learn
- Every map and model is an estimate. The useful question is how wrong it is, and where.
- Accuracy alone can mislead when classes are rare. Look at precision, recall and the confusion matrix.
Read →
Concept · Intermediate · 6 min read
How cameras capture 3D: optics, parallax and photogrammetry
From the pinhole camera to drone mapping: how photographs become measurements, and what limits their accuracy.
You will learn
- A camera turns 3D space into a flat image. Photogrammetry reverses that, using many overlapping photos.
- Ground sample distance depends on flying height, pixel size and focal length: GSD = height × pixel pitch ÷ focal length.
Read →
Concept · Beginner · 6 min read
The human body in VR and mixed reality: perception, latency and comfort
Why some virtual worlds feel natural and others make people ill, and the science of eyes, balance and timing behind it.
You will learn
- Comfort in VR is largely a matter of biology: vision, balance and timing must agree.
- Delays between head movement and the displayed image are a major cause of discomfort, so frame rate and latency matter.
Read →
Concept · Intermediate · 6 min read
How machine learning models learn: loss, gradients and why data matters
A plain-language tour of training, overfitting and generalisation, so you know what a model can and cannot be trusted to do.
You will learn
- A model is a function with adjustable settings. Training adjusts them to reduce a measure of error called the loss.
- Gradient descent improves the settings in small steps, like walking downhill in fog.
Read →
Tutorial · Beginner · 11 min read
Maths in code: distances on the Earth and rotations in 3D with Python
Put the fundamentals to work: measure real distances, rotate points, convert pixels to map coordinates and see floating-point limits for yourself.
You will learn
- The haversine formula measures great-circle distance and avoids the 11 percent error of treating degrees as flat.
- A rotation matrix preserves lengths, and the order in which rotations are combined changes the result.
Read →
Guide · Intermediate · 6 min read
How to check AI-generated GIS, 3D and machine learning code before you trust it
A practical review routine that catches the quiet mistakes, with the context to give an AI assistant so it gets things right the first time.
You will learn
- AI-generated code often looks correct and runs cleanly while being wrong in ways that only domain knowledge reveals.
- Give the assistant context: coordinate system, units, sensor, sizes and constraints, and ask it to state its assumptions.
Read →
When you finish
Where to go next
Next path
GeoAI: from zero to your first map
Learn the basics, work with free imagery and finish with a land cover map you tested honestly.
7 articles · 63 min
Next path
Immersive 3D, VR and AR
From choosing the right technology to capturing a real place, working with meshes and Gaussian splats, publishing it and training people with it.
13 articles · 86 min
Start a project
Want this applied to your own project?
Learning is the first step. If you would rather have a working system, book a discovery call and we will scope it with you.