The Ukrainian-Estonian company Farsight Vision has presented a major update to its neural network model for automated image analysis.
The company told this to Militarnyi.
The key feature of the new version is a significant expansion of the range of objects it can automatically detect.
While the previous model focused mainly on heavy equipment and medium-sized objects, the updated version can identify more than 30 types of objects, including small and hard-to-detect terrain features.
The detection function is already available on the FSV Platform. Access to object recognition is free for representatives of the Defense Forces of Ukraine.
The company said the updated model provides significantly higher detail and accuracy, enabling new use cases for reconnaissance and analysis of changes in photos and orthophotos.
The model can detect wheel tracks, narrow paths and concealed movement routes, as well as wires, antennas, poles and infrastructure elements.
It can also identify the locations of electronic warfare (EW) antennas and other camouflaged objects.
The update also expands supported data types. Users can already detect objects on orthophotos, and from May 15, 2026, the platform will also support analysis of standard photographs.
“We focus on reducing the time from receiving data to making a decision. The new model allows detection of changes that were previously overlooked and improves accuracy in difficult conditions,” said CEO Viktoriya Yaremchuk.
Farsight Vision is opening access for users to test the updated model and invites suggestions for additional object types to be included.
At the end of April, the company also launched the ability to stream video from a drone to the FSV Platform in real time during flight.
