MVTec D2S (MVTec Densely Segmented Supermarket)

MVTec D2S 是工业领域中实例感知语义分割的基准。它包含 21,000 张高分辨率图像，带有所有对象实例的像素标签。这些物品包括来自 60 个类别的杂货和日常用品。该基准的设计使其类似于自动结账、库存或仓库系统的真实设置。训练图像仅包含同质背景上的单个类别的对象，而验证集和测试集则更加复杂和多样化。

MVTec D2S is a benchmark for instance-aware semantic segmentation in the industrial domain. It contains 21,000 high-resolution images with pixel-level labels for all object instances. The objects cover grocery and everyday products from 60 categories. This benchmark is designed to emulate real-world scenarios such as self-checkout systems, inventory management, or warehouse operations. Training images only contain single-category objects on homogeneous backgrounds, while the validation and test sets are more complex and diverse.

搜集汇总

背景与挑战

背景概述

MVTec D2S是一个用于工业实例感知语义分割的基准数据集，包含21,000张高分辨率图像，涵盖60个类别的杂货和日常用品，并提供所有对象实例的像素级标签。其设计模拟自动结账、库存等真实超市场景，训练集图像背景同质且对象单一类别，而验证集和测试集则更加复杂和多样化，适用于评估分割算法在工业环境中的性能。