DSP Valley Newsletter
DSP Valley Newsletter
Easics’ deep learning live demo at the WoTS 2018 fair
(c) DSP Valley
From 2 till 5 October, easics participated in the World of Science and Technology exhibition in Utrecht, organized by FHI. At the World of Electronics section of the fair easics was proud to showcase its embedded deep learning demo for the first time to a broad audience.
The easics deep learning demo consists of an easics Daedalus board, a camera, and a display. Daedalus recognizes, localizes and tracks several classes of persons, animals and objects in real-time on live streaming video. Object classes include backpacks, laptops, bottles, cups, silverware, apples, scissors and umbrellas. The crowd was enthusiastic about the live demo.
Deep learning in a Box
Deep Learning in a Box is easics’ embedded AI (artificial intelligence) framework. It is deployed on-premise for real-time, low-latency inference, close to the sensors (as opposed to in the cloud). It operates in a small footprint, occupying only a small physical volume and consuming little power, compared to cloud-based solutions.
Compared to cloud-based solutions, Deep learning in a Box operates in a small footprint
Deep Learning in a Box relies on deep convolutional neural networks and easics’ automated code generation. It runs on several FPGA targets, complemented with state-of-the-art memories and connectivity. It can be deployed as a stand-alone box, as an FPGA SoM (system-on-module), or as an IP core in an FPGA. ASIC targets are being explored. Prototyping happens on an easics Daedalus board.
Applications of embedded AI
Easics focuses on embedded applications that benefit from object recognition, localization and tracking in images or live video. They include industrial machines, factory automation, robotics, autonomous vehicles, smart cameras, and crowd or traffic monitoring. Easics’ custom solutions rely on both AI and classical (non-AI based) techniques.
For an ever increasing number of applications, labeling data and training an appropriate deep convolutional neural network is significantly faster and cheaper than having an engineering team program the application using classical (non AI-based) machine vision techniques.
Embedding AI in your application
Applying Deep Learning in a Box to your application is fast, friendly and flexible. Contact easics to discuss your application and to see the easics deep learning demo.
Easics will verify the feasibility of applying Deep Learning in a Box to your application using the Deep Learning in a Box software tools.
The future of embedded AI
Want to learn more about the future of embedded AI, and easics’ take on it in particular? Then do attend the Leuven MindGate Visionary Seminar “Pushing the Limits of Artificial Intelligence” on Wednesday, November 28 in Leuven.