Vistas-NP

Vistas-NP 数据集是基于 Mapillary Vistas 数据集的分布外检测数据集。原始的 Vistas 数据集由 18,000 张训练图像和 2,000 张验证图像组成，共有 66 个类别。在 Vistas-NP 中，人类类被用作异常值，因为它们在场景中的分散以及与其他对象的视觉多样性。该数据集是通过将所有具有人员类别和三个骑手类别的图像排除到测试子集中来创建的。因此，该数据集有 8,003 个训练图像和 830 个验证图像。测试集包含 11,167 个。

The Vistas-NP dataset is an out-of-distribution detection dataset based on the Mapillary Vistas dataset. The original Vistas dataset consists of 18,000 training images and 2,000 validation images, with a total of 66 categories. In Vistas-NP, the human class is used as the outlier due to their scattered distribution in scenes and visual diversity relative to other objects. This dataset is created by assigning all images containing the person category and the three rider categories to the test split. Consequently, this dataset has 8,003 training images and 830 validation images, and the test set contains 11,167 images.

搜集汇总

背景与挑战

背景概述

Vistas-NP是一个基于Mapillary Vistas的分布外检测数据集，通过将人员类别作为异常值，从原始数据集中排除相关图像构建而成。它包含8,003张训练图像、830张验证图像和11,167张测试图像，主要用于图像分类、语义分割和异常值检测任务，由萨格勒布大学于2020年发布。