TinyML: Resource-Constrained Machine Learning

Student: Anthony Flores

Faculty Mentor: Suzanne Rivoire

Computer Science

College of Science, Technology, and Business

This presentation explores the comparative performance of machine learning on resource-constrained systems, specifically the Raspberry Pi 5 and XIAO ESP32S3. We aimed to quantify and compare the performance of image classification on each platform by collecting metrics on inference time, inference accuracy, and hardware utilization. By benchmarking identical image classification tasks across both platforms, our research examines the tradeoffs that generally come with implementing machine learning on resource-constrained hardware. Our results give insight into how these tradeoffs manifest in a machine learning context, which will serve to guide developers who are creating machine learning applications for embedded systems.