Automatic Field Boundary Detection Using Neural Networks

Agriculture of the 21st century is facing difficult challenges: feeding the growing population in a sustainable way is not going to be easy. We at EOSDA understand this problem and keep coming up with more and more advanced technological solutions to help growers, agri-businesses, retailers, and various institutions on a global scale.

We use high-resolution imagery and powerful boundary detection algorithms developed by the EOSDA’s own experienced team of scientists to achieve a high level of detail in field delineation. Our models can be adjusted to any territory by incorporating locally obtained data from the clients.

satellite field boundary map

How We Do It

over 80%

We offer an accuracy of above 80% depending on region, image characteristics, and season.

Minimum field size
3 ha

Minimum field size required for reliable boundary detection is 3 ha, but varies depending on the field's shape.

Data validation
IOU method

We use the intersection over union (IOU) method of validating the results, i.e. comparing the field boundaries we've detected with the data received from the client.


What You Get

Comprehensive Field Delineation

  • Get detailed field boundary maps for large areas that make crop classification much easier. We offer maps in .shp and .geotiff formats alike.

Field Delineation And API

  • Along with the field boundary maps, we offer fast and easy API access so you get all the information you need on particular fields of your clients.

High Accuracy

  • Reliable satellite imagery and innovative field boundary detection algorithms ensure the high accuracy of field delineation.

Stand-alone & Combined

  • We offer our field boundary detection as both a stand-alone solution and as the first stage for other EOSDA solutions, such as crop classification, advanced soil moisture analytics, yield prediction, and more.
boundary detection software

Advantages Of EOSDA Field Delineation


Field boundary detection is carried out remotely and at an instant.

Completely Automatic

No hands are required to carry out the field delineation; the algorithms do the whole job themselves.


The EOS SAT constellation will allow us to use the latest satellite imagery ensuring the field boundary maps are up-to-date.

Country-level Accuracy

High level of detail ensured even over large territories, such as a country.

Field-size Delineation

Our algorithms are capable of detecting the boundaries of even the smallest fields.

Change Detection

Detect changes in the field outlines by comparing the new boundaries with the existing databases.


We can train our field boundary algorithms for any location and resolution.


Our field boundary maps are in formats compatible with any other GIS software.

Ready Maps

Several field boundary maps are already available for you.


Field Delineation In Practice

Field boundary detection has many real-world applications:

For Governments
Automatic field boundary detection reveals the distribution of croplands over regions and countries in a very cost-efficient way.

Governments and administrative institutions can use this solution to determine subsidies, stay up-to-date about the state of crops in the area of interest, monitor production efficiency, and more.

There is also a sustainability angle here as the accurate field boundary maps allow for ensuring the successful implementation of new agricultural regulations that reduce the negative impact on the environment.

For Businesses
When it comes to running the agricultural business, information is king.

Accurate field boundary maps provide the foundation of large and detailed databases on various aspects of the crops growing in these fields, such as crop rotation, chlorophyll content, and more.

Crop classification, yield estimation, and harvest dynamics monitoring, and other solutions based on precision field delineation can empower decision-makers, helping them cut expenses and increase profits.


Success Stories

The EOSDA Crop Monitoring platform employs its own innovative field delineation method using convolutional neural networks (CNN) and artificial intelligence (AI).

The method allows us to: Automatically detect croplands. Outline farm boundaries on the map. Determine the acreage of the cultivated area. Obtaining this data significantly improves the accuracy of crop classification on a large scale.

Boundaries of Sugarcane Fields Detected in Brazil

Sugarcane fields in Brazil
Sugarcane fields in Brazil Green

Ukraine. Field Boundary Mask example.

field boundaries in Ukraine
Neural network crop classification
field boundaries in Ukraine refined
Refined results based on the field delineation