Get Started
The Agricultural Landscape Understanding API segments the landscape into different elements, leading to identification of different classes including individual field boundaries, trees, woodlands and minor irrigation structures like farm wells and dug ponds.
These enable insights at an individual field level and can form the fundamental base layer on top of which agricultural solutions are built, causing a shift from aggregate to individual level insights. The ability to share these insights dynamically for a specific latitude-longitude or an area of interest is the primary value proposition of the Agricultural Landscape Understanding output. The API provides an efficient and scalable way to provide the output of our Agricultural Landscape Understanding model
Understand The Landscape Segmentation Data
We solve the Agricultural Landscape Understanding problem using a multi-class panoptic segmentation model:
Data schema
Each landscape feature has an identifier which is basically a plus code of the centroid of the feature. (in extremely rare cases the identifier may not be unique).
API is designed to serve both researchers / academia as well as the needs of mobile applications for downstream on-ground stakeholders for agritech enterprises
Using the S2 library, the landscape data is sharded into small areas called S2Cells. These S2Cells are like little grids, and we used Level13, which makes them about 1 kilometer by 1 kilometer in size.
Landscape features are bucketed into s2cells by centroid location so a landscape feature will uniquely be only in a single cell