Efficient CNN Inference on Ultra-Low-Power MCUs via Saturation-Aware Convolution

Saturation-aware convolution reorders computation and checks for saturation to skip unnecessary work in quantized CNN inference while preserving output values.

Recommended citation: S. Li, L. Mottola, Yuan Yao and S. Kaxiras, "Efficient CNN Inference on Ultra-Low-Power MCUs via Saturation-Aware Convolution," 2026 Design, Automation & Test in Europe Conference (DATE), 2026.
Download Paper