Get Started
Home
Topics
Search
Library
Research questionHow can CNN inference fit on bare-metal microcontrollers with severe memory, binary-size, and runtime limits?Bare-metal microcontrollers may provide only a few kilobytes of SRAM and limited firmware space, while CNN inference also imposes execution-time demands. These constraints can make standard CNN deployment impractical even when the model's accuracy is acceptable.
AI
Computer Vision
Inference Optimization
Machine Learning
Research Paper
Small / On-device Models
Technology
Latest papersRecent research connected to this question, newest first.Deep Microcompression: Structured Pruning and Bit-packed Quantization for MicrocontrollersThe evidence covers LeNet-5 and deployments on the RP2040 (Cortex-M0+) and ATmega328P, including a dependency-free C library, reported weight compression, accuracy, binary-size, and deterministic-latency results. It does not establish performance across broader CNN architectures or microcontroller families.research paper · Sep 4, 2026
Related questions
How can depth compression merge padded convolutions without enlarging kernels for faster embedded inference?How can edge analog in-memory inference reduce energy while certifying errors from heterogeneous, imperfect accelerators?How can on-device LLM inference overlap weight I/O and computation under tight DRAM without stale sparsity decisions?How can passively cooled edge devices sustain DNN inference under thermal constraints?