msf-CNN: Patch-based Multi-Stage Fusion with Convolutional Neural Networks for TinyML
By: Zhaolan Huang, Emmanuel Baccelli
Potential Business Impact:
Makes tiny computers run AI using half the memory.
AI spans from large language models to tiny models running on microcontrollers (MCUs). Extremely memory-efficient model architectures are decisive to fit within an MCU's tiny memory budget e.g., 128kB of RAM. However, inference latency must remain small to fit real-time constraints. An approach to tackle this is patch-based fusion, which aims to optimize data flows across neural network layers. In this paper, we introduce msf-CNN, a novel technique that efficiently finds optimal fusion settings for convolutional neural networks (CNNs) by walking through the fusion solution space represented as a directed acyclic graph. Compared to previous work on CNN fusion for MCUs, msf-CNN identifies a wider set of solutions. We published an implementation of msf-CNN running on various microcontrollers (ARM Cortex-M, RISC-V, ESP32). We show that msf-CNN can achieve inference using 50% less RAM compared to the prior art (MCUNetV2 and StreamNet). We thus demonstrate how msf-CNN offers additional flexibility for system designers.
Similar Papers
On-Sensor Convolutional Neural Networks with Early-Exits
Machine Learning (CS)
Makes smart sensors use less power.
Efficient CNN Inference on Ultra-Low-Power MCUs via Saturation-Aware Convolution
Systems and Control
Saves power by skipping unneeded computer math.
Lightweight Software Kernels and Hardware Extensions for Efficient Sparse Deep Neural Networks on Microcontrollers
Machine Learning (CS)
Makes small computers run smart programs faster.