Last updated
3/4/2025
Get started
CompVis
Map Features
Map Features in OpenStreetMap with Computer Vision
This Blueprint helps you use object detection and image segmentation models to identify and map features in OpenStreetMap. It includes code for running inference on satellite images and exporting results for human verification and editing. Additionally, it provides notebooks for training custom object detection models on new datasets, allowing you to detect objects beyond the default swimming pool detection setup.
If you encounter any issues with the hosted demo below, try the Blueprint in the Google Colab Notebook available here.
Time
10 min
Complexity
Medium
Moderately complex. Some challenges but manageable
Preview this Blueprint in action
Hosted demo
Step by step walkthrough
Tools used to create
Trusted open source tools used for this Blueprint
Choices
Insights into our motivations and key technical decisions throughout the development process.
Focus
Decision
Rationale
Alternatives Considered
Trade-offs
Overall Motivation
Overall Motivation
Build a local-friendly AI helper for OpenStreetMap.
Enhances community-driven mapping with AI while keeping human verification.
Using an API-based solution that leverages Visual LLMs.
Using local-friendly and non-transformer models requires training data and finetuning step. Some Visual LLMs provided some outputs ‘Out of the box’, but the results were poor.
Object Detection Model
Object Detection Model
Finetuned YOLOv11 by Ultralytics.
State of the art, open source model, easy to finetune.
Out of the box pre-trained models.
Finetuning requires gathering and preparing a dataset.
Segmentation Model
Segmentation Model
SAM2 by Meta.
When provided with an input bounding box, there is no need to finetune to segment the object of interest.
Finetuning a custom segmentation model.
The dataset gathered for object detection is not “exploited” for segmentation.
Dataset Creation
Dataset Creation
Used OSM’s Overpass API & Mapbox Static Tiles API.
Open APIs that are flexible and can be easily queried.
Use an existing dataset with the objects of interest.
Using the APIs requires code to post-process the downloaded data.
Inference Pipeline
Inference Pipeline
Object Detection: Segmentation; Human verification.
2-step computer vision improves accuracy. Human verification ensures quality control.
Fully-automated. 1-step computer vision.
2-step computer vision pipeline introduces additional complexity and dependencies.
Deployment Options
Deployment Options
Supports Codespaces, Google Colab, HF Spaces.
Flexible for diverse environments and compute needs.
Single-path setups like only local.
Maintaining multiple pathways adds complexity for updates and consistency.
Ready? Try it yourself!
Discussion Points
Get involved in improving the Blueprint by visiting the GitHub Blueprint issues.
Join in
Explore Blueprints Extensions
See examples of extended blueprints unlocking new capabilities and adjusted configurations enabling tailored solutions—or try it yourself.
Load more
