Ternary-Input Binary-Weight CNN Accelerator Design for Miniature Object Classification System with Query-Driven Spatial DVS
This work addresses power and memory constraints for miniature imaging systems in space-constrained applications, representing an incremental improvement over prior CNN accelerators.
The paper tackled the problem of high energy demands in machine learning for miniature imaging systems by designing a CNN hardware accelerator that processes data from a spatial Dynamic Vision Sensor, achieving an 81% reduction in data size, 27% fewer MAC operations, and 1.6 mW power consumption with 440 ms inference time.
Miniature imaging systems are essential for space-constrained applications but are limited by memory and power constraints. While machine learning can reduce data size by extracting key features, its high energy demands often exceed the capacity of small batteries. This paper presents a CNN hardware accelerator optimized for object classification in miniature imaging systems. It processes data from a spatial Dynamic Vision Sensor (DVS), reconfigurable to a temporal DVS via pixel sharing, minimizing sensor area. By using ternary DVS outputs and a ternary-input, binary-weight neural network, the design reduces computation and memory needs. Fabricated in 28 nm CMOS, the accelerator cuts data size by 81% and MAC operations by 27%. It achieves 440 ms inference time at just 1.6 mW power consumption, improving the Figure-of-Merit (FoM) by 7.3x over prior CNN accelerators for miniature systems.