MMS-VPR: A Multimodal Street-Level Visual Place Recognition Dataset

MMS-VPR is a large-scale multimodal street-level visual place recognition dataset created by Yiwei-Ou. It comprises 110,529 images and 2,527 video clips with textual annotations, featuring day-night coverage and a 7-year temporal span in dense pedestrian-only environments. The dataset was last updated on HuggingFace on May 20, 2026.

Use Cases

• Train visual place recognition models based on multimodal street-level imagery and video.
• Benchmark model performance across day-night conditions using the dataset's temporal coverage.
• Develop multimodal AI systems that integrate visual and textual data for scene understanding.
• Evaluate the robustness of place recognition algorithms in dense, pedestrian-only environments.

Strengths

• Contains 110,529 images and 2,527 video clips, providing substantial scale for training.
• Features a 7-year temporal span and day-night coverage, enabling analysis of temporal and lighting variations.
• Integrates multiple data modalities including images, videos, and textual annotations.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.
• Data may reflect geographic or temporal bias inherent to its specific collection sources.

Provenance

Source

Yiwei-Ou on HuggingFace.

Time Range

7-year temporal span.

Freshness

Last updated 2026-05-20 10:00:56; freshness should be verified.

Geography

Dense pedestrian-only environments.