Summary of Hw-sw Optimization Of Dnns For Privacy-preserving People Counting on Low-resolution Infrared Arrays, by Matteo Risso et al.
HW-SW Optimization of DNNs for Privacy-preserving People Counting on Low-resolution Infrared Arrays
by Matteo Risso, Chen Xie, Francesco Daghero, Alessio Burrello, Seyedmorteza Mollaei, Marco Castellano, Enrico Macii, Massimo Poncino, Daniele Jahier Pagliari
First submitted to arxiv on: 2 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Hardware Architecture (cs.AR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A recent paper proposes an innovative approach for optimizing deep neural networks (DNNs) in low-resolution infrared array sensors. These sensors are crucial for people counting applications that prioritize privacy and energy efficiency. The authors develop a highly automated full-stack optimization flow, comprising neural architecture search, mixed-precision quantization, and post-processing. This comprehensive workflow enables the discovery of Pareto-optimal solutions balancing energy consumption, memory usage, and accuracy. By deploying these optimized DNNs on a customized hardware platform, the paper achieves significant improvements in model size reduction (up to 4.2x), code size reduction (up to 23.8x), and energy reduction (up to 15.38x) at iso-accuracy. This breakthrough has far-reaching implications for various applications, including occupancy monitoring and people flow analysis. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine a camera that can count how many people are in a room without showing their faces. This is possible with special sensors called infrared array sensors. These sensors are important for keeping our privacy while also saving energy. To make these sensors better, researchers developed a new way to optimize a type of computer program called deep neural networks (DNNs). Their approach combines different techniques to find the best balance between how well it works, how much memory it uses, and how much energy it consumes. By using this optimized DNN on a special device, they were able to make it smaller, more efficient, and use less energy while still working just as well. |
Keywords
* Artificial intelligence * Optimization * Precision * Quantization
