Automated classification of natural habitats using ground-level imagery.

Accurate classification of terrestrial habitats is critical for biodiversity conservation, ecological monitoring, and land use planning. Several habitat classification schemes are in use, typically based on analysis of satellite imagery and validation by field ecologists. Here, a methodology is presented for classification of habitats based solely on ground-level imagery (photographs), offering improved validation and enhanced ability to classify habitats at scale (e.g., using imagery from citizen science). In collaboration with Natural England, a public sector organisation with responsibility for nature/biodiversity conservation in England, this study develops a classification system that applies deep learning to ground-level habitat photographs, categorising each image into one of 16 distinct classes following the established 'Living England' framework. Images were pre-processed using resizing, normalisation, and augmentation techniques, while resampling was used to balance classes in the training data and enhance model robustness. A custom deep learning classifier based on the DeepLabV3-ResNet101 architecture was developed and fine-tuned to assign a habitat class label to ground-level photographs. Using five-fold cross-validation, the model demonstrated strong overall performance across 16 habitat classes, with accuracy and F1-scores varying between classes. This approach supports robust, scalable habitat classification based on balanced and well-prepared training data. Across all folds, the model achieved a mean F1-score of 0.63, with some habitat classes such as Bare Sand (BS) and Coniferous Woodland (CW) reaching values above 0.87. High performance was achieved for visually distinct habitats and lower performance for visually mixed or ambiguous classes. These findings demonstrate the potential of this approach for ecological monitoring. Ground-level imagery is easily obtained and accurate computational methods for habitat classification based on such data have many potential applications. To support use by practitioners, a simple web application is also provided that allows classification of uploaded images using the trained model.