Geoalert Blog

🛰️ My Imagery Search update in Mapflow Have you ever struggled to find the right image among all the files you’ve uploaded? A list of filenames is not always enough—especially when you need to check which images cover a specific project area without opening them one by one. Now you can search across your own imagery and collections using the same tools available in Mapflow’s Imagery Search: 📍 Search for your data geographically or by intersection with your area of interest 👉 Additionally you can fill out the metadata profile for your image which enables more filters like Date, Clouds, Off-nadir, etc. Your uploaded imagery is now easier to find, evaluate, and reuse across projects. ⚙️ Find suitable imagery and launch AI processing directly from the search results Try it in Mapflow → https://app.mapflow.ai

🎉Mapflow QGIS plugin 3.6.0 is out! 🛰️ Planned Imagery Search you already know from Mapflow Web — now lives right inside QGIS. If you regularly monitor the same locations, there’s no need to repeat the same imagery search every time. Define your areas once and let Mapflow watch for new satellite imagery. • Create a Planned Search and get notified when new satellite images are found over your areas • Add multiple named AOIs from a QGIS layer or simply draw them on the map • Check new imagery and launch AI processing directly from the search results • Keep processings organized by the AOI they belong to • Navigate the whole workflow — Planned Search → AOIs → processings — without leaving QGIS 🚀 Imagery Search, reworked • New Off-Nadir angle filter for sharper, more consistent imagery • All search now runs through the unified Mapflow catalog (legacy Maxar/Sentinel search removed) • Sort results server-side by clicking any column header • Use My Imagery — your own uploaded images and mosaics — as a search source 🚀 And more quality-of-life upgrades Update from the QGIS Plugin Manager and give it a try. Working with recurring monitoring or juggling many AOIs? Reach out — we're happy to help you automate it.

#Vegetation, tree canopy map. Australia 🇦🇺 User's #feedback.

🎉 New in #Mapflow: upload your own imagery straight through the Mapflow AI Agent No forms, no menus — just drop an image into the chat. The AI Agent now accepts image uploads right in the conversation. Drop a GeoTIFF into the chat window and the agent will: ✅ store it in your My Imagery collection ✅ select it as the input for your processing ✅ take you straight to running a model on it Bring your own aerial or satellite data and go from file to AI-mapping in one conversation. Contact us if you have questions or need support. 📖 How it works: https://docs.mapflow.ai/userguides/mapflow_agent.html 👉 Try it: https://app.mapflow.ai

A new #Mapflow Models Hub is live on our website. Explore production-ready #AI mapping models you can run today through the Mapflow web app, #QGIS plugin, or API. Can’t find the model you need? It probably means nobody has asked us for it yet—so let us know!

Tashkent is developing rapidly, with many modern business towers transforming the city’s skyline over the past two years. Uzbekistan has an ambitious urban development roadmap focused on expanding residential capacity, improving transportation systems, and allocating almost a quarter of the city’s area to green zones. A nationwide environmental campaign supporting these goals was launched in 2021. This makes it especially important to monitor urban development and make the results publicly available. While writing this short story about revisiting Tashkent during my business trip, my thoughts went back to the time when we worked to make useful datasets—such as the Tashkent vegetation dataset—openly accessible. Those datasets are still available on our GitHub. And the trees remain uncountable—because, unlike the new business centers, nobody is counting them.

Before we wish you a great and productive weekend, here's a quick reminder about the recent My Imagery update in #Mapflow. You can upload, organize, manage, and reuse your own imagery for AI processing. If you've already tried it, we'd love to hear your thoughts - check out this quick survey. 👈 Whether you've run into any issues or have ideas for improvement, even a short piece of feedback is incredibly valuable and helps us make Mapflow better. Have a great weekend! 🤗

🌲 Forest & Trees model — new version is live! The new v2026-07-03 is a major step forward for global vegetation mapping. We benchmarked it across areas on 4 continents where the previous version struggled — and the improvement is clear: 📈 F1: 0.513 → 0.850 📈 IoU: 0.396 → 0.744 📈 Recall: 0.407 → 0.868, with precision staying around 0.84 ✅ Higher F1 score in all 8 AOIs The previous model was very conservative and missed a lot of canopy. The new version captures sparse forest, shrubs, and tree lines much more completely — without adding extra false positives. 📊 We’re also publishing per-location benchmarks with side-by-side masks, and we’ll keep updating them. Try it in #Mapflow → https://app.mapflow.ai

Team balance management is now easier and less confusing. Team owners can assign credits to any team member - just edit. Please note that when a team is created, the full team balance is initially allocated to the Owner’s balance, so you’ll need to reallocate credits to other members. Read the full docs about Mapflow Team accounts

#Landuse, multi-model map. Indonesia 🇮🇩 User's #feedback.

At Geoalert we've been developing a product for Telco companies to optimise and design RF network, #RFN maps. Check the new article on our blog. "...continuous map maintenance can uncover thousands of missing buildings and major vegetation changes, making this approach a practical alternative to 1-3 yeras periodic map vendor's supplies"

🛰️ My Imagery major update in #Mapflow Web Until now, managing your own uploaded imagery in Mapflow meant using the #QGIS plugin or the API. You can now do it all right in the browser: 🖼️ Organize images into mosaics (collections) ⬆️ Drag & drop GeoTIFFs to upload 🖼️ Preview your imagery and check its status ⚙️ Start an AI-mapping run from any image in a couple of clicks If you already uploaded images before — nothing to re-upload. You can now drag images between mosaics to organise your collections. We call these collections "mosaics" because they can be processed as a single unit — for example, when you have multiple drone images that need to be analyzed together. Bring your own aerial or satellite data, run it through Mapflow's AI models for buildings, roads, forest and more, and keep everything reusable across projects. 👉 Try it: https://app.mapflow.ai/my-data 📖 Guide: https://docs.mapflow.ai/userguides/my_imagery.html

We posted about 🏠 Buildings model update and global benchmarks. Read on our blog 👀

🏠 Buildings model update and global benchmarks We’ve released the 2026 update of the Global Buildings model — one of the most important milestones for our R&D team. There are two main reasons that make us happy: first, this is our own a most demanded model in Mapflow and, second, it's aimed to work globally. In 2026, the 🏠 Buildings model is used in around 65% of all Mapflow processings. In 39% of these cases, users also enable the Height estimation option. The model is also part of our multi-model workflows, together with Forest and Roads, used to produce land-use-style maps. With this release, we have also started publishing our global benchmarks to make the evaluation process more transparent. The benchmarks are based on manually drawn building labels, cross-validated to reduce human errors. Some of these samples were collected from areas where previous model versions were less stable, including regions with complex and diverse urban morphology. 📊 Benchmark results The new Buildings v.2026–07–06 leads the previous production model on F1 in all 14 AOIs. On the global validation set of aerial imagery (9 AOIs) the mean area-based F1 rises from 0.849 to 0.893 (IoU 0.751 → 0.813); on the 5 dense-urban satellite imagery AOIs the mean F1 rises from 0.842 to 0.881 (IoU 0.729 → 0.787) The improvement is mainly driven by higher recall and precision in informal and high-density built-up areas such as in Saudi Arabia and India. In simple terms, the new model produces more detailed building outlines and is more stable across global urban patterns, thanks to larger training data and more diverse validation benchmarks. The important 🏠 Buildings model workflow update: We have also updated the simplification of the vectorized mask. It is now always applied by default. The updated method keeps the vectorized result closer to the original pixel mask and reduces annoying issues from previous methods, such as jagged edges and spikes. This release is another step toward reliable, practical, and globally scalable AI mapping. As always, we ask you to share your feedback. We review all ratings and try to get back whenever we can help. View more benchmarks in our documentation.

A huge forest map. 😳 How do you think how many individual tree crowns are there? #user_feedback, #Landuse, multi-model workflow, Italy 🇮🇹

New minor Mapflow QGIS plugin update 3.5.2 is out. The updates are small bug fixes and refactoring. However the #QGIS plugin remains one of the core user products and contunues be the primary source of credit consumption across the platform, supporting thousands of imagery analysis workflows directly from QGIS. Don't forget to update.

The #Mapflow AI Agent is available today in the Mapflow web app for all users. 🎉 • Open app.mapflow.ai, start a new project, and click Create project with AI Agent – the chatbox will pop up. • Or go to the old project and open chatbox to start the conversation. For the full workflow and the list of tools the agent uses, see the Mapflow AI Agent guide. 📊 Some statistics outcomes from the testing period As the most meaningful metric for this feature we chose project activation — the share of projects that have at least one processing run. It’s the moment a user stops exploring and actually runs a pipeline. 62% Agent beta users project activation rate 45.3% Mapflow platform average We’ll keep raising this number by adding new Agent skills, handling more edge cases, and fixing the errors. Thank you to everyone who tried it early. ❤️