HW-SW Optimization of DNNs for Privacy-preserving People Counting on Low-resolution Infrared Arrays
This work addresses the need for efficient and automated DNN optimization for people counting applications on resource-constrained hardware, representing a strong specific gain rather than a broad breakthrough.
The paper tackled the problem of optimizing deep neural networks for privacy-preserving people counting on low-resolution infrared arrays by proposing a full-stack optimization flow, resulting in up to 4.2x model size reduction, 23.8x code size reduction, and 15.38x energy reduction at iso-accuracy.
Low-resolution infrared (IR) array sensors enable people counting applications such as monitoring the occupancy of spaces and people flows while preserving privacy and minimizing energy consumption. Deep Neural Networks (DNNs) have been shown to be well-suited to process these sensor data in an accurate and efficient manner. Nevertheless, the space of DNNs' architectures is huge and its manual exploration is burdensome and often leads to sub-optimal solutions. To overcome this problem, in this work, we propose a highly automated full-stack optimization flow for DNNs that goes from neural architecture search, mixed-precision quantization, and post-processing, down to the realization of a new smart sensor prototype, including a Microcontroller with a customized instruction set. Integrating these cross-layer optimizations, we obtain a large set of Pareto-optimal solutions in the 3D-space of energy, memory, and accuracy. Deploying such solutions on our hardware platform, we improve the state-of-the-art achieving up to 4.2x model size reduction, 23.8x code size reduction, and 15.38x energy reduction at iso-accuracy.