Agenda: Vehicle Detection in Aerial Imagery for Domain Adaptation

Created by humansensinglab for ICCV 2025, this dataset facilitates the adaptation of vehicle detectors in aerial imagery to unseen domains using weak supervision. It integrates imagery and labels designed for domain-adaptation-object-detection tasks, specifically targeting remote sensing environments.

Use Cases

• Training vehicle detection models using weak supervision labels
• Benchmarking domain adaptation algorithms for aerial imagery
• Evaluating the impact of Stable Diffusion-generated data on object detection performance

Strengths

• Peer-reviewed research context from ICCV 2025
• Focuses on the niche domain of aerial vehicle detection
• Incorporates Stable Diffusion for domain adaptation

Limitations

• Unknown sample size and class distribution
• Potential label noise due to weak supervision methodology
• Lack of documented metadata for geographic or temporal context

Provenance

Source

humansensinglab, ICCV 2025

Collection Method

Weakly supervised annotation and Stable Diffusion generation

Freshness

Last updated December 2025.