Jan Erik van Woerden

An initial assessment of the suitability of VLM-based methods for V&V of defence applications is provided and a VLM-based error slice detection (ESD) method that independently groups and labels systematic errors made by a classification model is proposed.

Generative RGB-to-IR translation can boost infrared vehicle detection performance by over 10 mAP points in unseen UAV domains, challenging the notion that real data is irreplaceable.

Fine-tuning civilian models on military-relevant datasets can significantly boost target detection performance, but small object detection remains a critical hurdle.

Class-specific diffusion models can dramatically enhance military vehicle detection performance with as few as 8 training images, offering a game-changing approach in low-data scenarios.