tinyml

用户推荐 MIT 6.5940 Fall 2026 课程《TinyML and Efficient AI Computing》，由 Song Han 团队讲授，涵盖剪枝、量化、蒸馏、神经架构搜索、LLM 量化与部署、扩散模型优化等高效 AI 计算技术。

Lokutor团队发布ItoTTS，一个面向ESP32-S3芯片的流式英文语音合成引擎，每个语音仅4.89 MB权重、24 kHz音质，在自动自然度评测中接近StyleTTS 2水平，旨在为$5的边缘芯片上的本地LLM提供语音能力。代码以GPLv3开源，权重采用CC BY-NC-SA非商用许可。

MIT 6.5940: TinyML and Efficient AI Computing, taught by MIT Associate Professor Song Han, has open-sourced all its slides and videos — covering model compression, TinyML deployment, LLM and diffusion acceleration, and efficient training, with five hands-on labs.

ENAS is a hardware-aware Neural Architecture Search framework that operates efficiently on CPUs without GPUs, achieving speedups and competitive accuracy for TinyML models on resource-constrained microcontrollers.

The paper introduces TGL-NSGA-II, a teacher-guided fitness approximation framework for expensive evolutionary optimization in constrained TinyML neural architecture search, achieving improved efficiency and reliability over standard methods.

This paper introduces a streamlined pipeline for training and deploying machine learning models on the WeBe Band wearable device, focusing on system-level automation and hardware-aware optimization to enable rapid iteration and on-device evaluation for edge AI applications.

This paper presents ARGO, a smart eyewear platform for on-device machine learning that integrates a multimodal sensor suite and an optimized YOLOv11 model for real-time urban obstacle recognition, achieving privacy-preserving local processing with low latency and a memory footprint of 2.483 MB.

Proposes a lightweight neural architecture search performed directly on the deployment device for near-sensor computing, validated on sEMG sign language and fault diagnosis datasets, achieving improved accuracy and reduced RAM occupancy.

This paper presents an AI-driven framework for energy-efficient environmental monitoring in smart cities using edge intelligence and TinyML, which dynamically activates sensors based on spatiotemporal conditions to reduce energy consumption and extend sensor lifespan.