RIS-LAD
收藏资源简介:
RIS-LAD数据集是首个针对低空无人机场景的细粒度参考图像分割基准数据集。该数据集包含13871个精心标注的图像-文本-掩码三元组,从真实的无人机镜头中收集,重点关注小型、杂乱和多视角场景。该数据集突出了在先前基准中不存在的新挑战,例如由于小物体引起的类别漂移和在拥挤的同类别物体下的物体漂移。为了解决这些问题,我们提出了语义感知自适应推理网络(SAARN)。
The RIS-LAD dataset is the first fine-grained referring image segmentation benchmark dataset for low-altitude unmanned aerial vehicle (UAV) scenarios. This dataset contains 13,871 meticulously annotated image-text-mask triplets, collected from real UAV footage, with a focus on small-scale, cluttered, and multi-view scenarios. This dataset highlights new challenges absent in previous benchmarks, such as category drift caused by small objects and object drift under crowded same-category objects. To address these issues, we propose the Semantic-Aware Adaptive Reasoning Network (SAARN).

- 1RIS-LAD: A Benchmark and Model for Referring Low-Altitude Drone Image Segmentation厦门大学 · 2025年



