MMBench

MMBench是由上海交通大学开发的一个开源、端到端的多模态深度神经网络（DNN）基准套件。该数据集包含了一系列真实世界的多模态DNN工作负载，用于评估相关性能指标。MMBench覆盖了多媒体分析、情感计算、智能医疗、智能机器人和自动驾驶等多个重要研究领域，旨在通过详细的系统级和架构级分析，推动对多模态DNN工作负载的理解，并为未来的软件/硬件设计和优化提供洞察，以支持云和边缘计算平台上的多模态DNN应用。

MMBench is an open-source, end-to-end multimodal deep neural network (DNN) benchmark suite developed by Shanghai Jiao Tong University. This dataset contains a series of real-world multimodal DNN workloads for evaluating relevant performance metrics. MMBench covers multiple important research fields including multimedia analysis, affective computing, intelligent healthcare, intelligent robotics and autonomous driving. It aims to advance the understanding of multimodal DNN workloads via detailed system-level and architecture-level analyses, and provide insights for future software and hardware design and optimization to support multimodal DNN applications on cloud and edge computing platforms.