This week’s show focuses on news from Google I/O 2021 where it’s clear the search giant is trying to build out a fairly open ecosystem based on the Matter protocol and WebRTC for audio and video streaming. Google also said it would use Wear OS for Fitbit and combine Wear OS with Samsung’s Tizen OS. Google also announced a CarKey deal with BMW. There’s bad news on the data-sharing front from Imperial College London and Northeastern University, where researchers tracked how many connections popular smart home devices opened, and what it means for privacy. In security news, Consumer Reports found flaws in four security cameras and video doorbells, and we discuss the Eufy video camera bug. Additionally, cyber risks are so high that the CEO of Swiss Re, a reinsurer, said insurance for cyberattacks was becoming impossible. Finally, we mention the new Echo Frame options, although Kevin is still not a fan. In our hotline, we answer a question about Matter and keeping a SmartThings hub.
Our guest this week is Mark Hanson, VP of Innovation at Sony Semiconductor America. We talk about embedded computer vision and what it means to have machine learning taken care of on the image sensor itself. It enables lip-reading applications, occupancy sensing, and new ways to track inventory in stores. (He’s very excited about inventory sensing cameras.) Hanson also says the sensor and its DSP can provide training at the edge, allowing a user to show the sensor images and then have the sensor later recognize those images. We also talk about how product designers can figure out if they need a camera for a particular use case. Hanson really wants to get new ideas from everyone listening about use cases for embedded computer vision, so see if any of the interview sparks your creativity.
Hosts: Stacey Higginbotham and Kevin Tofel
Guest: Mark Hanson, VP of Innovation at Sony Semiconductor America
- Google will embrace Matter in most displays, smart speakers and on Android
- Do you know who your smart devices talk to?
- Cyberattacks are becoming too big for insurance to cover
- Why adding ML to an image sensor makes sense
- How to use a “dumb” sensor to offload some computing tasks
Podcast: Play in new window | Download | Embed