G-MAD: A Game-Based Data Generation Framework for Multi-View RGB-T Aerial Object Detection
G-MAD uses Arma3 to synthesize aligned multi-view visible and thermal aerial detection data with automatic bounding boxes.
The paper frames the tool as a way around costly real-world aerial dataset collection, weak viewpoint control, and imperfect RGB-T alignment. It supports scenario specification, controllable camera placement, simultaneous visible/thermal capture, and engine-derived annotation. The authors also release AMOD, a large-scale multi-view aerial RGB-T object detection benchmark built with the framework. HF Daily Papers' note
The paper frames the tool as a way around costly real-world aerial dataset collection, weak viewpoint control, and imperfect RGB-T alignment. It supports scenario specification, controllable camera placement, simultaneous visible/thermal capture, and engine-derived annotation. The authors also release AMOD, a large-scale multi-view aerial RGB-T object detection benchmark built with the framework. HF Daily Papers' note
score 4