Wake Vision
收藏资源简介:
Wake Vision数据集是一个用于TinyML系统研究的新颖数据集,由Technical University of Denmark和Microsoft共同创建。该数据集设计用于探索数据配置对TinyML系统性能的影响。数据集内容涵盖了不同时间和硬件约束下TinyML系统的表现,旨在解决如何通过数据感知优化提升TinyML系统效率的问题。尽管具体的数据量未在论文中明确,但该数据集被用于验证提出的新型数据感知神经网络架构搜索技术的有效性,并展示其在复杂数据集上的优势。
The Wake Vision dataset is a novel dataset for TinyML system research, jointly created by the Technical University of Denmark and Microsoft. This dataset is designed to explore the impact of data configurations on the performance of TinyML systems. It covers the performance of TinyML systems under varying temporal and hardware constraints, aiming to address the challenge of improving the efficiency of TinyML systems through data-aware optimization. Although the exact size of the dataset is not explicitly specified in the associated paper, it has been utilized to validate the effectiveness of the proposed novel data-aware neural architecture search technique, and to showcase its advantages on complex datasets.

- 1Fast Data Aware Neural Architecture Search via Supernet Accelerated EvaluationTechnical University of Denmark, Microsoft · 2025年



