Tibetan Plateau Poisonous Weed Segmentation Dataset (TPSD)
Dataset Summary

The Tibetan Plateau Poisonous Weed Segmentation Dataset (TPSD) is a fine-grained vision-language segmentation dataset constructed for poisonous weed identification in real-world field environments on the Tibetan Plateau. It is designed for referring image segmentation (RIS) and related vision-language grounding tasks in ecological monitoring.

TPSD contains:

field images collected from the Tibetan Plateau,

pixel-level instance masks for poisonous weeds,

semantically rich referring expressions corresponding to each annotated instance.

The dataset supports research on:

poisonous weed segmentation,

language-guided target grounding,

fine-grained ecological scene understanding,

cross-modal learning in complex vegetation backgrounds.