MCUBench
收藏资源简介:
MCUBench是由Deeplite, Inc.创建的一个综合基准数据集,包含了超过100个专为微控制器单元(MCU)设计的基于YOLO的对象检测模型。这些模型在Pascal VOC数据集上训练,并通过在四个不同的MCU平台上进行Pareto分析来选择。数据集的大小和复杂性使其适用于评估MCU在对象检测任务中的性能。创建过程中,模型在相同的训练管道下进行训练,确保了比较的公平性。MCUBench主要应用于边缘计算和TinyML领域,旨在解决在资源受限的MCU设备上高效部署对象检测模型的问题。
MCUBench is a comprehensive benchmark dataset created by Deeplite, Inc., which contains over 100 YOLO-based object detection models specifically designed for Microcontroller Units (MCUs). All these models are trained on the Pascal VOC dataset and selected via Pareto analysis across four distinct MCU platforms. The scale and complexity of this dataset make it suitable for evaluating the performance of MCUs in object detection tasks. During its creation, all models were trained under an identical training pipeline to ensure fair comparison. MCUBench is primarily applied in the fields of edge computing and TinyML, aiming to address the challenges of efficiently deploying object detection models on resource-constrained MCU devices.




