PrivateMap-Bench is a benchmark dataset for evaluating privacy leakage and utility loss in indoor Simultaneous Localization and Mapping (SLAM) map sharing. It contains sanitized SLAM artifacts, processed results, and executable code for reproducing the NeurIPS 2026 processed-artifact benchmark. The benchmark evaluates floor-plan, room-graph, trajectory, and routine leakage, as well as map compactness and navigation utility across 15 TurtleBot4 runs from three indoor layout variants.
Use Cases
- Benchmarking privacy-preserving SLAM algorithms based on the described leakage metrics (floor-plan, room-graph, trajectory, routine)
- Evaluating the trade-off between map utility and privacy based on the included aggregate privacy and utility results
- Reproducing academic benchmarks for indoor robotics based on the provided configuration files, code, and documentation
- Developing new privacy metrics for robotic map data based on the provided sanitized SLAM artifacts
Strengths
- Includes executable benchmark code and configuration files for full reproducibility
- Evaluates six specific metrics across 15 distinct robotic runs
- Contains sanitized processed artifacts, intentionally excluding raw, unreviewed, or private data
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
Provenance
- Source
- Subhash Chandra Dataverse
- Collection Method
- Likely generated from experiments with TurtleBot4 robots in three indoor layout variants.
- Freshness
- Last updated 2026-05-07 09:03:15; freshness should be verified