Deeplite has been on the road this past month! From March 22-25 we enjoyed ISC West in Las Vegas, checking out the latest and greatest in security and surveillance technology, and how computer vision AI could help accelerate current innovations. Then, from March 28-30 the team exhibited at the tinyML Summit in Burlingame, California. There we met with the growing computer vision AI ecosystem, listened to interesting talks about how others in the space are using tinyML, and chatted with partners about the work we’re doing to improve AI for embedded devices.
Here are a few takeaways from both shows.
This was Deeplite’s first time attending ISC West, which is the premier security and surveillance tradeshow in North America. We were happy to see so many top companies at the show – the big brands as well as smaller startups. Overall, there was a real buzz and feeling of excitement in the air as everyone could be together in person again, to showcase new products and innovative applications. We spent our time visiting with companies to discuss how Deeplite and its work with computer vision could help accelerate their AI adoption in security cameras and associated services. We’re now looking forward to ISC East later this year!
A common theme at ISC West was reducing power consumption, and the need to do more with less. The cost of hardware, like most things nowadays, is expensive and prices are still rising. Companies are looking for cheaper ways of doing things, while using less power and creating less of a carbon footprint, as being “greener” was also top of mind.
While at the show, we spoke to a number of security and surveillance companies, including those that deploy security cameras in homes and businesses. Deeplite is well positioned to optimize devices like these, as they are often low on space and compute power. Additionally, now that companies are offering more in terms of analytics, they are facing challenges regarding compute power for their installed cameras – again, the need to do more. One example we saw was a company whose hardware was detecting people, but now they want to offer their customers additional services. This could include recognition of clothing color, and/or supporting a multi-class detection model for vehicles, packages, pets and people. But, to be able to deliver this on smaller processors with power efficiency, high speed and accuracy is proving to be a challenge. This is exactly where Deeplite can help, with our unique software solution and expertise for enabling computer vision AI on embedded devices.
Exhibitors checking out the booths at ISC West
Another prominent theme at the show was companies wanting to run computer vision AI more cost effectively, by transitioning from cloud to edge-based video analytics. The main driver was certainly cost savings, but quality-of-service and data privacy and security were also of high importance. It was great to see many companies promoting some kind of AI and/or digitization strategy, which we saw as another trend. Clearly these companies are adopting AI to further their societal innovations and we were able to talk with many about how Deeplite could help bring it to the next level, in a cost-efficient manner.
We also spoke with several companies at the tinyML Summit, which is another compelling event bringing all the brightest minds in tinyML together - where we were thrilled to be back with all the participants in this community. Deeplite is a big supporter of advancing tinyML education and adoption – our co-founder and chief product officer, Davis Sawyer, previously wrote about this in an article for Unite.Ai. Being at the show gave us a great opportunity to discuss some of the challenges we’re all facing, particularly the need to do more with less.
This was Deeplite’s third time supporting the tinyML Summit, this year as an Executive Sponsor. This year’s show was very technical in terms of the audience and content, with a lot of people giving talks about how they are using tinyML in the community for good. We met with deep learning and AI engineering colleagues, as well as researchers. There were also a lot of vendors in the microcontroller and sensor space – companies driving a very small (tiny) footprint.
As an Executive Sponsor, Deeplite attended an advisory board meeting during the summit, where we discussed the importance of data, and its quality. It’s no secret that data is at the center of everything we do. In order for AI to work, we need a lot of data to train it properly. This plays nicely with what we do for our customers – such as labeling and training data as part of an efficient AI pipeline. Because data can take up a lot of space, and with the need to limit power consumption, organizations like Deeplite are essential to optimize AI inference to perform more tasks.
Members of the Deeplite team at our tinyML Summit booth
We were fortunate to hear from some very interesting speakers at the summit, and how they are using tinyML to do good in the world. The talks were split into different sections: Vision, Audio, Sensors, Hardware and Software, and it was interesting to see different companies’ takes on how they are using tinyML within these categories. One such speaker was Kate Kallot, who is a Director at NVIDIA and chair of the tinyML for Good Committee. She showed us some inspiring videos and discussed how she’s been spending time in parts of Africa, using tinyML to help kids, especially young girls, obtain access to educational content. Additionally, she discussed how tinyML and data are making a positive social impact, such as in combatting climate change, making clean water accessible and contributing to farming, especially in these remote areas. It’s these kinds of stories that make us proud and excited about what we do, and what AI can accomplish.
Overall, it was just a great atmosphere to see everyone again in person and to connect with some of our hardware partners, like Arm and Alif, among others. We’re looking forward to the next shows and events that we will be attending, including the Embedded Vision Summit in Santa Clara next month.