Diagnosing Aerial-View Object Detectors with Foundational Image Generative Models
For researchers and practitioners in aerial object detection, this provides a method to diagnose model weaknesses and efficiently improve performance, though the approach is domain-specific and incremental.
This work introduces a synthetic diagnostic framework for aerial-view vehicle detection that uses text-guided generation and attribute-controlled editing to create a controllable testbed. The framework reveals performance gaps in pretrained detectors under diverse conditions, and targeted data supplementation guided by these diagnostics improves AP50 by up to 13% with fewer samples than non-targeted augmentation.
Recent advances in large-scale image generative models enable photorealistic scene synthesis with controllable attributes. Beyond data augmentation, their potential as diagnostic tools for trained vision systems remains unexplored in the aerial and remote sensing domains. We introduce a synthetic diagnostic framework for aerial-view vehicle detection that combines text-guided generation, attribute-controlled editing, and automated attribute verification to construct a controllable synthetic testbed. This enables fine-grained evaluation of pretrained detectors under diverse scene types and environmental conditions that are difficult to isolate in real datasets. Across three detection architectures and three real aerial datasets, synthetic scene-wise performance trends closely match real-world weaknesses. Guided by these diagnostics, targeted supplementation with small real datasets from the identified weak categories yields improvements of up to 13% AP50 while requiring substantially fewer additional samples than non-targeted augmentation. Our results show that controlled synthetic probing can predict real-domain performance gaps and guide efficient data collection. The proposed diagnostic framework is modular and can incorporate alternative generative or vision-language models as capabilities evolve. Our code and datasets are available here: https://humansensinglab.github.io/AVODDiag/