Target-PLP is a cross-domain test-only dataset for zero-shot evaluation of aerial power-line detection models.
It contains 194 aerial frames annotated with polyline instance labels in YOLinO format. Models are trained on TTPLA and evaluated on Target-PLP without fine-tuning.
Compared to TTPLA, Target-PLP is more complex and challenging: denser wire layouts, more parallel instances per frame, greater scene variation, and a different imaging domain. Annotations are… See the full description on the dataset page:
https://huggingface.co/datasets/V8heart/Target-PLP.
