uTensor/uTensor

Health v2 · maintenance only

Adoption not part of the score

1928 stars · 250 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

uTensor is an extremely lightweight machine learning inference framework built on TensorFlow and optimized for Arm Cortex-M microcontroller targets, with a core runtime of only ~2KB. It converts trained TensorFlow models into generated C++11 code plus a small runtime library for on-device inference.

Use cases

• run neural network inference on a Cortex-M microcontroller
• deploy a TensorFlow model to embedded hardware
• run TinyML on an IoT device with limited RAM
• generate C++ code from a trained TensorFlow model
• do on-device ML inference without a cloud connection

When to choose

• you need ML inference on resource-constrained Arm Cortex-M or mbed targets
• you want a minimal runtime footprint (~2KB) and predictable memory usage
• you prefer offline model translation to generated C++ code

When to avoid

• you need training on-device rather than inference only
• you target non-Arm platforms or need broad operator support
• you want an actively evolving ecosystem like TFLite Micro

Facets

library · maturity maintenance

machine-learning llm-inference embedded machine-learning embedded-systems iot cpp embedded cross-platform tinyml tensorflow cortex-m edge-computing microcontroller mbed model-codegen

Member repositories

For agents

markdown · JSON · MCP: product_card(name="uTensor/uTensor")