HyperCam: Low-Power Onboard Computer Vision for IoT Cameras

2025-01-17

AI Summary

Resource-constrained IoT cameras demand ultra-lightweight, energy-efficient vision models deployable on microcontroller units (MCUs), yet existing deep learning and classical ML approaches incur prohibitive memory, latency, or accuracy trade-offs. Method: This work introduces the first end-side lightweight visual recognition framework for MCU-class IoT cameras based on Hyperdimensional Computing (HDC). It proposes a full-stack lightweight training–inference pipeline integrating HDC-based modeling, MCU-optimized operators, and custom wireless camera hardware. Contribution/Results: Evaluated on MNIST, Fashion-MNIST, and face detection tasks, the framework achieves 72.79%–93.60% accuracy with only 0.08–0.27 s inference latency and minimal memory footprint—42.91–63.00 KB Flash and 22.25 KB RAM—outperforming SVM and MobileNetV3 baselines. This is the first systematic application of HDC to embedded vision systems, enabling unprecedented co-optimization of accuracy, energy efficiency, and resource utilization.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Large Vision Models
• Humans and AI: Human-in-the-loop Machine Learning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

We present HyperCam, an energy-efficient image classification pipeline that enables computer vision tasks onboard low-power IoT camera systems. HyperCam leverages hyperdimensional computing to perform training and inference efficiently on low-power microcontrollers. We implement a low-power wireless camera platform using off-the-shelf hardware and demonstrate that HyperCam can achieve an accuracy of 93.60%, 84.06%, 92.98%, and 72.79% for MNIST, Fashion-MNIST, Face Detection, and Face Identification tasks, respectively, while significantly outperforming other classifiers in resource efficiency. Specifically, it delivers inference latency of 0.08-0.27s while using 42.91-63.00KB flash memory and 22.25KB RAM at peak. Among other machine learning classifiers such as SVM, xgBoost, MicroNets, MobileNetV3, and MCUNetV3, HyperCam is the only classifier that achieves competitive accuracy while maintaining competitive memory footprint and inference latency that meets the resource requirements of low-power camera systems.

Problem

• IoT Cameras
• Energy-efficient Computer Vision
• Low-power Image Recognition

Innovation

• HyperCam
• Energy-efficient
• Computer Vision