ForestPersons: 100K+ RGB and Thermal Images for Forest Search-and-Rescue

ETRI released ForestPersons in 2026, a collection of 100,000 to 1,000,000 images for search-and-rescue person detection in forest environments. The data includes RGB and thermal imagery captured from ground and low-altitude drone perspectives featuring subjects in various poses and visibility levels.

Use Cases

• Training object detection models to locate persons at 20% visibility in dense foliage
• Developing sensor fusion models using paired RGB and IR thermal imagery
• Optimizing UAV-based detection algorithms for low-altitude perspectives

Strengths

• Contains between 100,000 and 1,000,000 records
• Dual-modality coverage with RGB and Thermal (IR) sensors
• Four specific visibility levels (20, 40, 70, 100) for testing occlusion

Limitations

• Simulated scenarios may lack the environmental unpredictability of actual search-and-rescue operations
• Geographic bias toward US forest environments

Provenance

Source

ETRI (Electronics and Telecommunications Research Institute), ICLR 2026

Collection Method

Sensor-based capture using RGB and IR cameras on ground and UAV platforms

Freshness

Last updated February 2026.

Geography

United States