Bringing Artificial Intelligence to the Edge

Bringing Artificial Intelligence to the Edge obviously injects new possibilities to the world of Internet of Things. Things get smarter, can draw conclusions locally and in turn, become more useful. One of the applications of AI on the Edge which excites me the most is Computer Vision.

Project State

Public Project

Licences

Software Licence: Apache 2.0
Hardware Licence: Project has no hardware

Project Tags

Admins

byteLAKE

Does this project pique your interest?

Login or register to join or follow this project.

Bringing Artificial Intelligence (AI) to the Edge obviously injects new possibilities to the world of Internet of Things (IoT). Things get smarter, can draw conclusions locally and in turn, become more useful. One of the applications of AI on the Edge which excites us at byteLAKE the most is Computer Vision (CV).

Computer Vision is all about converting the video streams into valuable and useful information. The way it works is that the computer analyzes the pixels from the consecutive video frames, and generates information about what's happening. Eventually, we get information about demographics, activities, we can count the objects or people in particular. We can track the objects, analyze their behaviors, look, emotions and many more. We can also analyze how behaviors change over time and for instance answer the question of: how long people spent at certain locations, what were they doing there etc. All this information can be later used as a foundation for building a comprehensive analytics system helping for instance retailers optimize their operations.

Computer Vision, however, comes with a cost. It relies on Deep Learning algorithms which are extremely demanding and do not necessarily fit into the ecosystem of constrained devices. So how can we still have low-power and high-performance under one roof? How can we bring Computer Vision to the world of Edge devices? Let we briefly explain how we tackled on that issue.

We have built a computer vision system based on  Raspberry Pi 3 , low power high performance Intel® Movidius™ Myriad™ 2 Vision Processing Unit (VPU) and Basler's camera of the following configuration: camera daA1600-60uc (2MP, 1600 px x 1200 px, 60pfs) + lens Evetar M12B0416IR F1.6 f4mm 1/2". It also includes byteLAKE's Computer Vision asynchronous model to enable real-time on-device objects classification. It has been designed to make it applicable across wide selection of industries, as part of i.e. intelligent security cameras, gesture controlled drones, industrial machine vision equipment, and more. In the video below, you can see an example application for the Retail industry:

``

 

If you are interested in enhancing your products or services with Machine Learning or Computer Vision capabilities, get in touch with me directly or contact byteLAKE's team. We believe that the above solution will significantly reduce the time-to-market for intelligent IoT deployments and in turn translate into reduced costs of such efforts.

Comments

Your comments, please!

Want to comment this ... Show more