Thermal UAV Imagery for Cross-Modal Geo-Localization

Thermal-UAV is a dataset for cross-modal thermal geo-localization, designed for Unmanned Aerial Vehicles in GNSS-denied environments. It addresses the modality gap between thermal queries and visible-light satellite imagery. The dataset was created by FloralHercules and was last updated in April 2026.

Use Cases

• Train models to match thermal UAV query images to visible-light satellite reference imagery for geo-localization.
• Develop semantic segmentation models on thermal imagery to identify landmarks for consensus-based localization frameworks.
• Benchmark cross-modal retrieval performance using thermal image features against a database of geo-tagged satellite images.

Strengths

• Dataset is introduced in a peer-reviewed research paper (SCC-Loc), providing a defined benchmark.
• Specifically designed to address the challenging problem of cross-modal geo-localization for UAVs.

Limitations

• Specific quantitative details like image count, resolution, geographic coverage, and class balance are unknown.
• The dataset's primary application is a niche research problem, limiting generalizability to other vision tasks.

Provenance

Source

Introduced in the paper 'SCC-Loc: A Unified Semantic Cascade Consensus Framework for UAV Thermal Geo-Localization'.

Collection Method

Likely collected using UAVs equipped with thermal cameras and paired with satellite imagery; exact methodology is unspecified.

Freshness

Last updated in April 2026, indicating recent maintenance.