SynthSoM
收藏资源简介:
SynthSoM数据集是由北京大学等研究机构开发的用于机器联觉(SoM)研究的合成智能多模态感知-通信数据集。该数据集涵盖了多种空-地多链路协作场景,包含多种天气条件、时间、智能体密度、频段和天线类型。数据集包含射频通信数据(如373K组信道矩阵和104K组路径损耗)、毫米波雷达感知数据(如230K组雷达波形和61K组雷达点云)以及非射频感知数据(如455K张RGB图像、891K张深度图和241K组LiDAR点云)。数据集的创建基于一个集成了AirSim、WaveFarer和Wireless InSite的仿真平台,并通过基于统计的定性检查和机器学习评估指标进行了验证。该数据集旨在为SoM相关算法的交叉比较、模型校准和基线实现提供一致的数据支持。
The SynthSoM dataset is a synthetic intelligent multimodal perception-communication dataset developed by Peking University and other research institutions for machine synesthesia (SoM) research. This dataset covers various air-ground multi-link collaboration scenarios, and includes diverse weather conditions, times, agent densities, frequency bands and antenna types. The dataset contains radio frequency (RF) communication data (e.g., 373K channel matrix sets and 104K path loss sets), millimeter-wave radar perception data (e.g., 230K radar waveform sets and 61K radar point cloud sets), and non-RF perception data (e.g., 455K RGB images, 891K depth maps and 241K LiDAR point cloud sets). The dataset was created based on a simulation platform integrating AirSim, WaveFarer and Wireless InSite, and validated through statistical-based qualitative checks and machine learning evaluation metrics. This dataset aims to provide consistent data support for cross-comparison of SoM-related algorithms, model calibration and baseline implementation.

- 1SynthSoM: A synthetic intelligent multi-modal sensing-communication dataset for Synesthesia of Machines (SoM)北京大学先进光通信系统与网络国家重点实验室, 山东大学-南洋理工大学人工智能联合研究中心, 东南大学国家移动通信研究实验室 · 2025年



