Google DeepMind and WRI release 10-meter global map classifying six forest and land-use types

·2 views

Google DeepMind, Google Research, the World Resources Institute and IIASA released Forest Typology 2020 (ForTy), a global map classifying land into six forest and land-use types at 10-meter resolution. The classes follow FAO definitions, including primary, naturally regenerating, planted and plantation forest, tree crops and agroforestry, and other land. The classifier uses AlphaEarth Foundations embeddings and was trained on about 1.7 million labeled locations, reporting 90.2 percent overall accuracy against 8,190 expert-labeled sites. The dataset, described in Scientific Data, is free on Google Earth Engine under an open license and includes a time series of yearly forest change from 2017 to 2025. It was designed as a reference layer for the EU Deforestation Regulation's 31 December 2020 baseline.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Guide Explains Using Security Headers to Shield Production Websites

A technical guide on DEV Community explains how security headers can strengthen production websites against common attacks. It highlights Content Security Policy for limiting permitted content sources, which helps block cross-site scripting. It also covers HTTP Strict Transport Security for forcing encrypted HTTPS connections and secure cookie flags that restrict JavaScript access to sensitive cookies. The article adds best practices such as applying headers across all production sites, reviewing them regularly, and testing them with online validation tools.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Stray zero-width spaces in Android app translations mostly enter via repositories, not Weblate

A census of F-Droid app translations found that Weblate-hosted apps contain about a quarter as many stray zero-width spaces as other apps. Tracing 90 such characters in 23 Weblate apps showed none were added by a Weblate commit, with most arriving alongside new strings or languages. Sources included translation commits, syncs from Localazy, Crowdin and Transifex, and code copied from other projects. A control check confirmed the detection method could catch Weblate-saved characters, and the authors report that predictions about six languages where the character is legitimately used largely failed. The authors publish these results as negative findings and suggest flagging the 'space plus two zero-width spaces' pattern, while noting the evidence is thin in those languages.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Gamma vs Canva for AI Presentations in 2026: Which Suits Makers Best

The article compares Gamma and Canva Magic Studio for creating AI-assisted presentations in 2026, noting they share a feature category but serve different purposes. Gamma is presented as the better choice when the main challenge is quickly producing a coherent first draft from a brief, outline, or PDF, and when a shareable web link is acceptable as the deliverable. Canva is recommended when presentations are part of a broader brand system that includes social media, ads, and one-pagers, and when pixel-level control and Brand Kits matter. The author suggests many freelancers and consultants may use both tools, starting in Gamma for structure and moving to Canva for multi-asset brand consistency. The piece also outlines a basic workflow for Gamma and cautions that exports may not perfectly preserve certain design effects.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

New Free Tool Strips Hidden Location and Device Data From Photos Without Uploading Them

A developer has released Drop Metadata, a free tool that removes hidden metadata such as GPS coordinates, device details, timestamps, and camera serial numbers from photos and videos. Phone images often carry this embedded information, which can reveal where a person took a picture, even when the image itself looks harmless. Unlike many similar services, the tool processes files entirely on the user's device, so nothing is sent to a server. It removes metadata by rewriting only the metadata sections, leaving the image or video data unchanged and avoiding quality loss. The tool supports JPG, PNG, MP4, and MOV files, and an Android version works offline.

0 comments Read more at DEV Community