Geoalert Blog

A sneak peek at the Mapflow Web redesign 👀 Coming sooooon! Ready to dive into a whole new user experience? 🤯️️️️️️

We’re sharing a new customer story from Brazil 🇧🇷 about Celplan’s switch from manual digitization to an AI mapping workflow for telecom clients. This shift helped them increase mapping productivity by 83–160%, depending on the area. Thank you, Celplan, for sharing your experience! We hope this is just the beginning of our journey together.

The production of our maps used to rely on manual building vectorization and estimated building heights, resulting in time-consuming processes, limited scalability, and high operational costs. By integrating Mapflow into our workflow, we have been able to accelerate map production, improve efficiency, and enhance scalability while reducing our reliance on manual processes. The support and collaboration of the Mapflow team have been essential throughout this process, particularly in customizing the AI model to meet our specific needs and technical requirements

From the user's #feedback. Classification of the industrial buildings and warehouses. Philippines 🇵🇭

A main contributor to #Mapflow Agent just recently proposed a new term: "vibe-mapping". 😅

#Landuse, multi-model map. Egypt 🇪🇬, Cairo. Mapflow User's #feedback.

From car segmentation to a parking plan A Mapflow user told us how their CAD team draws parking layouts by hand, bay by bay, for unlabeled lots. So we prototyped a simple app. If an aerial image has enough cars parked in it, the cars themselves show where the bays (car spaces) are — spacing gives the width, orientation gives the rows. Enough to draft a layout plan and skip the tedious part. As an innnput data we used Mapflow car detections. The point isn’t only the detector. The whole workflow is projected to run through Mapflow MCP, so you can point an agent at one parking area or a thousand to scale it up with the platform. Try it on your own image, or one from Mapflow’s connected sources. Just make sure there are enough cars. Give us some ideas of what to prototype next. We will share back or even provide you with the free credits to prototype on your own. 🙂

From terrain to a radio plan, then kept current 1. Load the terrain Start from a terrain model of the area: Copernicus GLO-30 or the elevation data you already use. 2. Generate the city from fresh imagery Mapflow classifies land use and extracts every building and tree, with its height, from recent satellite imagery. 3. Plan the radio network Load the layers into your planning tool and predict coverage on the city as it stands today. (We convert data into Atoll's or other software formats) 4. Update only what changed When new imagery shows a change, such as this construction site, Mapflow re-maps only that area and you recalculate the network. 👉 Learn more about Mapflow Telecom tailored solutions. 👉 Request us to run a demo or pilot project in the area of your network.

🚗 Mapflow Roads Model — Update! We've released a significantly improved version of our Roads segmentation model! Roads is one of Mapflow's core basemap models, primarily designed to extract narrow roads and local road networks, especially in areas where OpenStreetMap and other public maps have limited coverage. 📊 How much better? We evaluated the new model using SpaceNet 3 benchmark datasets and our validation data. Across seven datasets, compared with the previous one: • IoU: +55% — more accurate road coverage • Topo-F1: +63% — better road network connectivity The second metric is especially important for routing, where roads may be partially hidden by trees, buildings or shadows. 🗺️ Two workflows available: • 🚗 Roads — connected road networks, postprocessed into buffered polygons for easier mapping. • Buildings + Vegetation + Roads — lighter road postprocessing, better suited for combined land-cover analysis. 👉 Give it a try and rate your processing results

🚀 AOI list management further improvements While you're viewing the processing results ("View on the map") you can edit AOI names, iterate through the list of AOIs and download each area's geometry and results. Especially useful when you process many AOIs in a single run.

🧑‍🤝‍🧑Team management, redesigned To tell the truth it didn't go well with the user's adoption of the first version becuase of UI, it's appeared to be more than confusing. Teams in Mapflow are now much simpler. • "Team" has moved from Settings to the main sidebar menu • The Admin invites people by email, and they join with one click • Members spend the Team's credits, while their personal balance stays untouched • The Admin adds or withdraws credits for any member right from the members list 👀 Both your Team and personal balances are shown in the header 👉 Guide docs

🌲 Forest and trees workflow was updated with two models, both benchmarked against thoroughful manual labels. 🚀 Tree crowns, v.2026-09-10 Crown polygons re-evaluated on 7 AOIs against manually annotated crowns: mean F1 0.824 → 0.849 Neighbouring canopies are separated instead of merged — which is what tree counting actually depends on. Run it with the Tree crowns option on 0.3 m imagery (zoom 19): polygons, or points. 🚀 Canopy Height Model, v.2026-08-31 Benchmarked against lidar-derived ground truth on 5 m hexagons. Across 38 AOIs the mean absolute height error drops from 4.70 m to 3.24 m (−31%). Better heights mean a more reliable split into shrubs <4 m / forest 4–10 m / forest >10 m — the ranges used for vegetation management along power lines and transport corridors, which is customizable. 📖 Read more in our docs: 📊 Per-location benchmarks, with prediction-vs-ground-truth overlays: 1) Crowns 👈 2) CHM 👈

📍AOI management update Have a long list of areas to map? You can process them one by one, or combine them in a single GeoJSON file and run them all in one go. Now you can easily manage your areas of interest (AOIs) directly in the map view: navigate the list, check results, and download AOI geometries. See how it works in a short video. 👀🤗

🌳 How many trees are in New York’s Central Park—and what can we learn from imagery? The source of truth, Central Park Conservancy, cites more than 18,000 trees. We explored the park with Mapflow’s updated vegetation and tree crown detection model, which mapped 23,846 crowns. This is an automated detection result, with differences from the inventory still to investigate. The workflow started with a single prompt to Mapflow Agent. 👨‍🔧 From mapping trees and estimating their heights to detecting cars or solar panels, describe your task and let the Agent select the relevant model, run the processing pipeline, and help interpret the results. What would you map in your area?

We are sharing a story from our long-term client Instituto de Información e investigación geográfica, estadística y catastral del Estado de México (IGECEM 🇲🇽). Francisco Polo, Geography Department Manager:

IGECEM uses Mapflow to streamline urban area monitoring for cadastral purposes. As cities grow, IGECEM leverages Mapflow custom model to track urban sprawl. By analyzing changes in building coverage over time the Institute can gain valuable insights into city development patterns.

Multi-model, Landuse + Results summary. Go to demo processings in Mapflow public project to test it.

📊 #Mapflow now summarizes your processing results Getting a vector layer is only part of the imagery analysis workflow. The next question is: what did the model actually find? Now every Mapflow result opens with an instant summary: 🏠 Buildings — counts and types 🚗 Roads — total length 🌍 Land Use — area by category Each model highlights the statistics that matter for its output. You can download the results as a CSV file, or ask the Mapflow #Agent about them directly. The Agent receives the structured statistics instantly and can help you understand the results in more detail. The new service is powered by Statistics API and available for every default model. Custom-model workflows currently include basic statistics, but their summaries can also be customized. PDF reports will come soon. We love this new feature and hope you will enjoy it. 🤗 Tell us which metrics you would like to see in a custom or default workflow's summary.

We’re sharing a new story on the Mapflow blog: a geographical snapshot of Mapflow demand in 2026 so far. On the interactive map, you can switch between the number of jobs, average area per processing, and total area covered. So far this year, 6,521 people have run 18,745 processing jobs, mapping 44,433 km². That’s roughly the land area of Estonia—mapped one shot at a time, by users across 178 countries. So, what does the map reveal?

Afer-coworking time on the stairs of Konzerthaus, Berlin. Occasional #workspace. 👨‍💻😎

A brief line-up of AI-mapping #Mapflow models. The models are the core stack of the platform that orchestrates them and provides services to manage the input data. Mapflow turns satellite and aerial imagery into ready-to-use vector maps. Here's what's in the box: 🏠 Buildings — global, 0.3 m / z19. Extracts rooftops, plus optional building-type classification, contour regularization, simplification, merge with OSM, and height estimation (beta) that projects the roof down to ground level for true 3D footprints. 📊 v.2026-07-06: mean F1 0.893 (aerial), 0.881 (satellite) 🌲 Forest & trees — global, 0.6–0.3 m. A solid vegetation mask covering sparse forest, shrubland, small tree groups and narrow tree lines. Options: height classes (<4 m / 4–10 m / >10 m) and individual tree crowns as polygons or points. 📊 v.2026-07-03: mean F1 0.850 — up from 0.513 in the previous version 🚗 Roads — 0.3–0.5 m, tuned for rural and suburban areas. Multi-task learning keeps the mask connected where roads hide under trees; centerline extraction plus graph cleanup (gap merging, simplification, removal of stray segments). 🧩 Buildings + Roads + Forest — all three in a single run, returned as one topology-corrected GeoJSON. 🗺 Open Data — pull OpenStreetMap and Overture layers with no AI involved, as a complement to model output. ⚙️ Custom models, on request: construction sites, high-density housing, agriculture fields, Segment Anything (zoom-dependent, from land use down to single trees), swimming pools (F1 > 0.95), solar panels. You can also bring your own model — or ask us to train one for your use case. Run it all in the web app, the #QGIS plugin, or via #API. 🔗 mapflow.ai/models · docs.mapflow.ai

🛰️ 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