FreeLLMGoogle DeepMind releases ForTy, a 10‑m global forest‑type map for 2020
The ForTy map classifies land into six forest and crop categories with 90.2% overall accuracy, aiding EU deforestation regulations.
Google DeepMind and the World Resources Institute unveiled ForTy, an AI‑driven global map of land cover for the year 2020.
The map uses AlphaEarth Foundations and three image‑recognition stages to assign each 10‑meter pixel to one of six classes, including primary forest, plantations and tree‑crop farms.
Independent validation of 8,190 sites showed a 90.2% overall correctness rate, with 89% accuracy for primary forests but lower performance for plantations and tree crops.
ForTy outperformed comparable European Commission datasets, scoring 95.5% versus 93.6% for forest‑versus‑non‑forest and 90.9% versus 87.8% for forest‑type classification on the same test data.
The map provides a 2020 reference point for the EU Regulation on Deforestation‑free Products, supporting supply‑chain risk assessments and broader conservation goals.
あなたにとっての意味
The dataset gives policymakers and companies a detailed baseline for evaluating land‑use changes since the 2020 cut‑off, improving compliance with the EUDR. Researchers and developers of free or local AI tools can use the ForTy data via the GeeFlow library to train or test geospatial models without needing proprietary satellite pipelines.
ソース
- Google Deepmind creates a global map of forests with satellite data and AI, The Decoder, 31 Oct 2026
- GitHub - google-deepmind/geeflow: GeeFlow - generate and process large-scale geospatial datasets with Google Earth Engine. · GitHub, github.com (official), 31 Oct 2026
上記のソースから自動的に生成し、当社のコードでチェックしました：すべての数値と名前が含まれ、文がソースをコピーしてはいけません。間違いを見つけましたか？報告する。
今日無料で使えるAIはどれですか？
すべての無料AIモデルをブラインドヒューマン投票で順位付けし、あなたのGPUで必要なメモリと速度を示します。