Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Battery-Free, Connected Machine Vision
Saman Naderiparizi, Joshua Smith, Matthai Philipose
University of Washington
Microsoft Research
Aaron N. Parks, Vamsi Talla, Bryce Kellog, Benjamin Ransford, Zerina Kapetanovic, Yi Zhao, James Youngquist, Pengyu Zhang, Seungyeop Han, Bodhi Priyantha, Jie Liu, Deepak
Ganesan, Jason Goldstein, Shyam Gollakota
Why Battery-Free Sensing?
• By 2020 there will be 50 billion connected devices and 44ZB data generated per year.• 50 billion dead batteries?!
• Internet of battery-free things
Battery-Free Camera
• One of the most challenging sensors to wirelessly power• High power consumption
• Rich output data
Battery-Free Camera Limitation
1) On-board Storage
2) Computational Capabilities
3) Communication Speed
Wireless Power System
Power and Communication
On-Board Computing
Power and Data Interface
Off-Board Computing for High-Level Control and Heavy Compute Tasks
Computation Partitioning Example
Impossible Possible Possible
Face Recognition on Battery-Free Camera
Low-Power Wearable Continuous Mobile Vision
Glimpse Camera, Programmable early discard camera architecture:• Most pixel are irrelevant• Some low-power sensors can predict them
inertial/light
thermal
grayscale pair
blurry/
dark?
body temp?
< 1.5m?
discard
discard
discard
windows
windows
primary imagers
Algorithm Development
FIR Tracking
Stereo Pair
Summary
• WISPCam a battery-free camera.
• Implementing the interfaces to connect WISPCam to plugged in PC.
• Pushing the envelope further to bring machine vision into battery-free domain.
Thank you!Questions?
sensor.cs.washington.edu
mailto:[email protected]