Map-Free Reloc: Metric Pose Estimation Benchmark for ECCV 2022

Niantic Labs published this benchmark in 2022 for the ECCV paper 'Map-free Visual Relocalization,' which focuses on estimating metric pose relative to a single reference image. The dataset provides a standardized evaluation framework for visual relocalization tasks that do not rely on pre-existing 3D maps or scene-specific training. It includes image pairs and ground truth 6DoF relative poses to facilitate research in map-free navigation.

Use Cases

• Estimating 6DoF relative pose between a query image and a single reference image
• Benchmarking feature matching algorithms using metric pose accuracy metrics
• Training PyTorch-based neural networks for map-free visual localization

Strengths

• Peer-reviewed ECCV 2022 benchmark
• Includes ground truth 6DoF metric pose labels
• Developed by Niantic Labs research team

Limitations

• Narrowly focused on single-image relocalization rather than full SLAM
• Repository lacks explicit documentation on the total number of scenes or images

Provenance

Source

Niantic Labs

Collection Method

Annotated sensor data from visual relocalization research

Freshness

Last updated February 2026; based on 2022 ECCV research.