This paper proposes TinyDef-DETR, a DETR-based framework for automatic transmission line fault detection using drone imagery. TinyDef-DETR integrates an edge-enhanced ResNet backbone that enhances edge-sensitive representations, a stride-free space-to-depth module that preserves detail, a cross-stage dual-domain multi-scale attention mechanism that jointly models global context and local cues, and a Focaler-Wise-SIoU regression loss function that improves the localization of small, challenging targets. Extensive experiments on public and real-world datasets demonstrate that TinyDef-DETR effectively mitigates the limitations of existing detectors, achieving excellent detection performance and generalization ability while maintaining reasonable computational overhead. It is particularly suitable for UAV-based transmission line fault detection, especially in scenarios involving small, ambiguous targets.
