DeepTrees Germany: Modelled Tree Segmentations from Digital Orthophotos (DOP) by Federal State

This repository contains high-resolution, multi-temporal tree crown segmentation datasets across German Federal States, generated using DeepTrees—an open-source deep learning framework based on PyTorch. Utilizing 20 cm-resolution DOP imagery, we provide pixel-wise instance masks and comprehensive ecological metrics (e.g., crown area, crown spraed, vegetation indices) for millions of trees. Data is provided as geojson files per DOP tile per year per state and supports applications in urban forestry, biodiversity monitoring, green infrastructure planning, and environmental research. Here is an overview table: Federal State -- Initial Years Dataset Available Sachsen -- SN 2023 Yes Sachsen Anhalt -- ST 2022 Yes Baden-Württemberg -- BW No Brandenburg -- BB No NIedersachsen -- NI No Saarland -- SL No Thuringia -- TH No Berlin -- BE No Rhineland-Pfalz -- RP No Nordrhein-Westfalen -- NW No Hamburg -- HH No Schlieswig-Holstein -- SH No Mecklenburg-Vorpommern -- MV No Hesse -- HE No Bremen -- HB No Bayern -- BY No