Mapillary

Mapillary
Name	Mapillary
Founded	2013
Founder	Jakob Wrazidlo; Jan Erik Solem
Headquarters	Malmö, Sweden
Industry	Geospatial, Photogrammetry, Computer Vision
Parent	Meta Platforms (2020–present)

Mapillary is a collaborative platform for street-level imagery and geospatial data aggregation that enabled crowdsourced image capture, photogrammetric processing, and map improvement. Founded in 2013 by entrepreneurs with backgrounds in robotics and computer vision, the service grew into a global repository of panoramic and directional images used by researchers, municipal agencies, and private companies. Mapillary combined mobile capture apps, web-based management, and machine learning pipelines to extract features such as signs, road markings, and building facades for integration with mapping projects and urban planning.

History

Mapillary was founded in 2013 amid growing interest in crowdsourced mapping exemplified by projects such as OpenStreetMap, Wikimedia Commons, and initiatives by Google like Google Street View. Early fundraising and incubation drew attention from investors linked to Y Combinator and European startup networks in Stockholm and San Francisco. The company partnered with civic actors in cities including Malmö, New York City, São Paulo, and London to pilot street-level imagery initiatives aligned with municipal open data goals associated with entities such as European Union urban programs. Strategic collaborations included technology firms like Esri and transport agencies such as Transport for London for pilot integrations. In 2017–2019 Mapillary expanded its developer ecosystem through APIs and SDKs, reflecting trends from platforms like Mapbox and HERE Technologies. In 2020, Mapillary was acquired by Facebook, Inc. (later Meta Platforms), joining other imaging and mapping efforts linked to companies such as Instagram and WhatsApp and prompting discussions with advocacy groups in United States and European Commission policy circles. Post-acquisition, Mapillary’s datasets continued to interoperate with open mapping communities including OpenStreetMap, and its machine learning models contributed to research at institutions like MIT, Stanford University, University of Oxford, and corporate labs such as DeepMind.

Platform and Features

The platform combined mobile applications for capture with cloud-hosted processing, offering features comparable to services from Google, Apple, Bing Maps, and HERE Technologies. Core components included a smartphone app compatible with iOS and Android, support for dedicated camera rigs used in projects by companies like Bosch and research teams at ETH Zurich, and server-side photogrammetry similar to systems developed by Trimble and Autodesk. Mapillary’s web interface provided mapping overlays interoperable with Leaflet, OpenLayers, and Mapbox GL JS for visualization, while an API allowed integration with GIS packages such as QGIS and ArcGIS. Computer vision pipelines employed convolutional neural networks inspired by architectures from ImageNet challenges, with object detection and semantic segmentation approaches related to work at Google Research, Facebook AI Research, and Microsoft Research. Features included sequence-based image organization, automatic GPS trajectory smoothing akin to techniques in Garmin devices, 3D point cloud generation comparable to outputs from LiDAR systems, and vector extraction for elements similar to datasets maintained by TomTom.

Data Collection and Workflow

Data collection workflows emphasized crowdsourcing and institutional capture. Individual contributors used mobile apps to upload geotagged photos, echoing models utilized by Flickr and Panoramio contributors. Municipal or corporate partners deployed vehicle-mounted rigs referencing sensor stacks employed by autonomous vehicle developers like Waymo and Cruise LLC for higher-density coverage. Uploaded imagery entered a processing pipeline involving metadata normalization, deduplication, and feature extraction using machine learning models trained on annotated datasets from projects such as COCO and Cityscapes. Workflow integrations enabled batch uploads via desktop tools and continuous feeds from fleet operators, akin to telemetry ingestion systems at Uber and FedEx. Versioning and change detection capabilities facilitated temporal analyses similar to those used by remote sensing programs from European Space Agency and NASA.

Licensing and Open Data Policies

Mapillary navigated tensions between proprietary corporate ownership and community-oriented open data practices. Initially, community contributions were governed by licensing choices designed to be compatible with open mapping initiatives like OpenStreetMap and licensing frameworks such as those promulgated by Creative Commons. Following corporate acquisition, policy discussions referenced precedent cases involving Wikimedia Foundation and licensing disputes in the tech sector involving Google and Microsoft. Mapillary provided export options and terms of use that affected downstream usage by governments, NGOs such as Humanitarian OpenStreetMap Team, and companies like Esri and TomTom. Data governance debates engaged standards organizations including Open Geospatial Consortium and regulators like the European Data Protection Board over data portability and interoperability.

Use Cases and Applications

Applications spanned urban planning, transportation engineering, asset management, and academic research. City agencies in municipalities such as Copenhagen, Barcelona, and Vancouver used imagery to audit infrastructure conditions in projects comparable to initiatives led by UN-Habitat and World Bank urban resilience programs. Transportation agencies and consultants from firms like AECOM and Arup leveraged extracted road and sign inventories for traffic modeling akin to work by Institute of Transportation Engineers. Companies in insurance and real estate used imagery for risk assessment and property documentation similar to practices by Allianz and Zillow. Academic groups at Carnegie Mellon University and Imperial College London applied the dataset for machine learning research, benchmarking against datasets such as KITTI and Oxford RobotCar. Humanitarian actors used imagery for crisis mapping tasks coordinated with United Nations Office for the Coordination of Humanitarian Affairs and volunteer organizations like MapAction.

Privacy and Ethical Concerns

Privacy debates mirrored controversies around street imaging by Google Street View and surveillance worries raised in policy forums involving European Commission, Federal Trade Commission, and civil liberties organizations such as Electronic Frontier Foundation and Privacy International. Key concerns included biometric data capture of individuals, automatic plate recognition parallels with law enforcement tools used by agencies like Interpol, and implications for sensitive sites referenced in discussions with UNESCO and urban planners. Mitigation measures included automated blurring technologies inspired by academic work at University of Cambridge and legal compliance processes responding to statutes like the General Data Protection Regulation in European Union. Ethical discourse involved scholars from institutions such as Harvard University, University of California, Berkeley, and advocacy groups including Access Now debating consent, data stewardship, and equitable access.

Category:Geographic information systems