Multimodal Manufacturing Dataset

该数据集包含了来自火箭装配线原型的一系列测量数据，其中包括两台摄像机同步拍摄的图像以及多种传感器测量数据。具体来说，数据集包含了传感器变量、输送带变频、驱动器温度、机器人物理属性、输送带工作站统计、循环状态、循环计数、异常类型，以及两台摄像机对应的图像文件名等测量信息。该数据集跨越了285个装配周期，共有166,000条记录。其任务是对装配线中的异常进行预测。

This dataset contains a series of measurement data from a rocket assembly line prototype, including synchronously captured images from two cameras and various sensor measurement data. Specifically, the dataset includes sensor variables, conveyor belt frequency conversion settings, drive temperatures, robot physical attributes, conveyor belt workstation statistics, cycle status, cycle count, anomaly types, as well as image file names corresponding to the two cameras and other measurement information. This dataset covers 285 assembly cycles and contains a total of 166,000 records. Its task is to predict anomalies in the assembly line.

搜集汇总

背景与挑战

背景概述

该数据集是'NSF-MAP'项目的一部分，专注于制造装配线中的异常预测，包含从'Future Factories(FF) Dataset'预处理得到的多模态和模拟数据。数据集支持多种基线模型和提出的融合方法，旨在提高异常预测的鲁棒性和可解释性。