ADE-OoD
收藏arXiv2025-09-30 收录
下载链接：
https://ade-ood.github.io/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为ADE-OoD基准，是基于ADE20k数据集构建的，包含了来自不同领域、具有高语义多样性的图像。这些图像涵盖了室内外场景，并含有多种分布外的物体。在ADE-OoD上，DOoD方法的表现超越了先前的方法，但仍有改进的空间。该数据集具有高语义多样性，包含了150个语义类别，其任务是针对语义分割进行分布外检测。
The dataset in question is the ADE-OoD benchmark, constructed based on the ADE20k dataset, which encompasses images from diverse domains with high semantic diversity. These images cover both indoor and outdoor scenes and contain various out-of-distribution (OOD) objects. On the ADE-OoD benchmark, the DOoD method outperforms prior state-of-the-art approaches, yet there remains room for further improvement. This benchmark features high semantic diversity, includes 150 semantic categories, and its targeted task is out-of-distribution detection for semantic segmentation.
