Mapillary Training Datasets - AI at Meta

2025-07-21 · Source: ai.meta.com via Google News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Robotics & Autonomous Systems · Depth: Intermediate, quick

Summary

Mapillary offers a suite of training datasets derived from over two billion street-level images contributed globally via smartphones and action cameras. These datasets support the development of recognition models by providing diverse benchmarking data, including sequences, depth data, outdoor imagery, city-level imagery, and semantic segmentations. Key datasets include Vistas, CrowdDriven, Metropolis, Planet-Scale Depth, Street-Level Sequences, and Traffic Sign. Established in 2013 and acquired by Facebook (now Meta) in 2020, Mapillary continues to provide public access to street-level imagery and map data, leveraging computer vision and machine learning to extract map features for various mapping projects.

Key takeaway

For research scientists developing computer vision models, Mapillary's extensive and diverse street-level datasets offer a critical resource for training and benchmarking. You should explore specific datasets like Vistas or Planet-Scale Depth to enhance model robustness and accuracy, particularly for urban mapping, autonomous driving, or environmental analysis applications.

Key insights

Mapillary provides diverse, global street-level image datasets for training computer vision models.

Principles

• Crowdsourcing enhances data diversity
• Computer vision extracts map features

Method

Mapillary collects street-level images from a global network of contributors, then uses computer vision to extract map data and create diverse training datasets for recognition models.

In practice

• Train recognition models on street-level data
• Utilize semantic segmentations for mapping
• Access depth data for 3D applications

Topics

• Mapillary
• Training Datasets
• Street-Level Imagery
• Computer Vision
• Semantic Segmentation

Best for: Research Scientist, Computer Vision Engineer, Machine Learning Engineer, AI Scientist

Related on AIssential

• Building and Road Recognition in Dense Urban Informal Settlements: A Dataset and Benchmark — Existing remote sensing datasets and models are insufficient for accurate infrastructure mapping in dense urban informal settlements.
• What is Semantic Segmentation? A Key Technology Powering Modern AI Vision — Semantic segmentation classifies every image pixel, enabling precise AI understanding of visual environments.
• I Taught a Computer to Read Indian Roads. Here Is What Nobody Tells You About That Problem. — Autonomous driving models require adaptation to unstructured road environments beyond Western standards.
• Google's new AI tools put film scouting in Street View and promise to cut weeks of satellite analysis to minutes — Google's new AI imaging tools streamline creative integration, urban planning analysis, and infrastructure identification.
• 346 Blog Posts To Learn About Computer Vision — Computer Vision leverages AI to interpret visual data, enabling diverse applications from image generation to real-world object detection.
• UrbanWell: Benchmarking Multimodal Large Language Models for Spatio-Temporal Urban Wellbeing Analytics — UrbanWell benchmarks MLLMs for spatio-temporal urban wellbeing, revealing performance variability across diverse indicators.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by ai.meta.com via Google News.