Artificial Intelligence (AI) / Deep Learning / Convolutional Neural Networks (CNNs)

The majority of AI development is currently happening in the cloud. The latest trend is a move for AI from the cloud to the edge. Embedded deep learning is a solution when low latency is key and has a need for small, power efficient and low-cost devices.    

Easics embedded AI solutions is covering this new market segment. It focuses on local, real-time and low-latency processing inside the machine. Our solution works on-premise (close to the sensors) and in a small footprint at low cost and low power consumption. Within this embedded space, we will focus on applications benefiting from image recognition.

The following markets are benefiting from our AI solutions: industry 4.0, agriculture, healthcare, automotive and space.

Easics deep learning platform is based on convolutional neural networks. This  framework is currently running on FPGA, and we want to explore ASIC targets as well.

Deep learning is popular in image processing and computer vision for:

  • object classification: what
  • object detection: what & where
  • object tracking: object detection over a time series

You can start your deep learning journey here, with the following steps: 

  • Step 1: gather data and label them 
  • Step 2: select a neural network or model 
  • Step 3: train your model 
  • Step 4: embed the inference engine in your application

deep learning in a box

Yolo v3 demo

Daedalus board

    Easics has built a prototype AI platform for object recognition and detection using a deep learning inference engine on FPGA

    Easics proposes different system approaches to embed deep learning in your product or application:

    • IP core on FPGA or CPU 
    • IP Core on System on Module (SoM)
    • Deep learning in a box, connect your machine via ethernet and make it Artificial Intelligent.

      Easics deep learning solution is fast, friendly and flexible and has the following benefits for you:

      We match performance with cost

      • frame rate

      • resolution

      • high bandwidth and low latency

      • power budget

      • size (volume)     

      We implement the interfaces of your choice e.g ethernet, PCI-express, ... 

      Flexible programmable accelerators and hardware images of different models available

      Future-proof software stack ownership still adaptable to the fast-moving AI space and evolving model types

      Turnkey deployment of convolutional neural networks (CNNs)

      Integration is possible with your existing hardware or new developments

      Long-term support and scalability

      Applications that can benefit from embedded AI can be found in industry 4.0, agriculture, healthcare, automotive and space: 

      • Application-specific machine vision systems
      • Smart cameras
      • Robotics 
      • Industrial machines
      • Automation
      • Intelligent traffic systems
      • Quality control in Semiconductor, chemical or pharmaceutical production

      Deep learning in a box