Article on how microcontrollers in resource-constrained environments work! Performing ML inference on tiny hardware improves privacy by running locally, lowers costs for model vendors, and eliminates dependence on network. Notes anil.recoil.org/notes/sigmob..., article dl.acm.org/doi/10.1145/...
GetMOBILE article on benchmarking micro-NPUs now out
Resource-constrained, MCU-scale neural processing units have emerged to provide real-time inference within milliwatt-scale power budgets.
anil.recoil.org