Automatic outlier removal
Cleans dense data without manual rules
Each point is compared with its neighborhood, the right way to filter yield monitor and high-density sensor data.
Removes invalid data, global outliers and local spatial outliers.
Geospatial processing for precision agriculture
Turn field data into precision maps. In one call.
Reliable, with no manual tuning. Ready for APIs, the cloud and AI agents.
pip install apbase- 01 Smart
Removes local outliers and picks the best mathematical model on its own.
- 02 High performance
A Fortran core with OpenMP, running on every CPU core.
- 03 Scalable
From notebooks to plugins, GIS, APIs and cloud clusters.
Smart
Raw field data in. Precision map out.
Yield monitor data, soil samples and sensor grids arrive irregular and noisy. APbase decides on its own what to clean and which model to use.
Best mathematical model
Picks the interpolation with the lowest error
Compares models with cross-validation and uses the one with the lowest error on your data. No manual tuning.
Model A
Model BChosen
Model C
- 01Filters outliers
- 02Builds boundary and grid
- 03Picks the model
- 04Delivers the clipped map
Use APbase with
Python and Rust
One call turns your field points into a precision map. The same core runs in Python and Rust, and comes to other languages next.
Language roadmap
- PythonAvailable now
- RustAvailable now
- ReactUp next
- GoUp next
main.py
Read the Python docsimport apbase# OpenMP parallelism: set once per processapbase.config["n_threads"] = 4# Filters, builds the grid and picks the best modelresult = apbase.create_map(x, y, z, resolution=10.0)result.method # model chosen by cross-validationresult.z # finished map, clipped to the boundary
Read the Rust docsuse apbase::MapOptions;// OpenMP parallelism: set once per processapbase::config.set("n_threads", 4)?;// Filters, builds the grid and picks the best modellet result = apbase::create_map(&x, &y, &z, MapOptions::new(10.0))?;result.method; // model chosen by cross-validationresult.z; // finished map, clipped to the boundary
High performance and scalable
A fast engine that scales with your application
The engine
- 01
Fortran + OpenMP
Native kernels on every core, bundled with the package.
- 02
Lightweight, no dependencies
Takes coordinates and values and returns the map. No web framework, no database.
- 03
Deterministic
Same input, same map, every time.
- Notebook
- Plugins and GISIntegration with GIS software
- API
- Workers and queues
- ContainersKubernetes, Cloud Run, ECS
- AI agentsMCP servers
Same pipeline, same result at every level