Landscape Observatory is developing a new digital annotation tool that will significantly accelerate the production of the high-quality training data needed to support the creation of AI maps across England’s Protected Landscapes.
As the project prepares to scale map production across 25% of England, one of the big challenges is the volume of aerial photography that must be manually interpreted and annotated to train the machine learning models.
The new annotation tool has been designed to support the GIS & Digitisation Technical Team by streamlining this process, enabling digitisers to work more efficiently while maintaining the consistency and accuracy needed to develop reliable AI models.
Tackling the Challenge of Scale
Artificial intelligence can only identify land cover as accurately as the data it has been trained on. Before AI can automatically classify landscapes, thousands of aerial images must first be manually interpreted by trained specialists who identify and label different components of the photograph.
Producing this training data is one of the most time-intensive stages of AI development.
The new annotation tool has been designed to reduce this burden by making the manual annotation process faster, more consistent and easier to quality assure. This will enable the team to process significantly larger volumes of aerial imagery as AI map production expands.
From Annotation to AI
In late July, members of the GIS & Digitisation Technical Team spent three days at Cranfield University working alongside remote sensing experts Dr Dan Simms and Dr Toby Waine to continue the development and testing of the annotation tool.
The visit provided an opportunity to stress test the web application, created by Senior GIS Developer Ann Holden, and understand the complete AI development pipeline from manual annotation through to model training and validation.
For many of the graduates, this was the first opportunity to see how the data they produce directly influences AI model performance. Experiencing the full workflow strengthened their understanding of how careful annotation, quality assurance and habitat verification contribute to accurate land cover mapping.
Supporting the Future of AI Habitat Mapping
As Landscape Observatory scales up AI map production, digital tools such as the annotation application will play an increasingly important role in improving efficiency while maintaining scientific rigour.
By reducing the time required for manual annotation and supporting the production of consistent, high-quality training data, the application will help accelerate the development of AI models capable of mapping habitats across England’s Protected Landscapes (covering all National Parks, National Landscapes and National Trails).
The continued partnership with Cranfield University ensures that the latest advances in remote sensing, machine learning and geospatial science are translated into practical tools that support better environmental monitoring, informed decision-making and nature recovery.
The development of the annotation tool marks another important step towards delivering AI-enabled habitat mapping at a national scale, providing the evidence needed to better understand, manage and protect our landscapes for the future.