Low Power CV - Industrial Automation, Low power microcontroller based ML

06 June 2021 - less than 1 min read time
Tags: AI Low Power CV ML

Low Power CV - Industrial Automation, Low power microcontroller based ML

Low Power CV

What is Low Power CV ?

Edge or Mobile Computer Vision
Microcontroller based Computer Vision

Edge or Mobile Computer Vision

Frameworks: Linux, Assembly Level, RISC-OS (V)

Power Ranges: 5V - 24 V DC , 1-8 A

Libraries and Deployment: Most of the deep learning libraries and deployment frameworks

Programming Languages : Python, C++

Use of Onboard Edge Devices in multiple fields now a days is most common.
Some of the famous edge devices are:
1. Jetson Nano
2. Jetson Xavier AGX
3. Jetson Xavier NX
4. Google CoralTPU USB Adapter
5. Google Coral Board
6. Raspberry Pi
7. Beagle Bone
8. Panda Board
9. Asus Boards etc.
Edge Devices can be connected to SDKs and can be used in different deployment methods to optimise costs.
- AWS GreenGrass
- GCP IOT
- Azure IOT
- With the help of open-source SDK’s
  - Some of them are:
    1. Kafka
    2. Kubernetes
    3. ROS

Deployment Frameworks:

Tensorflow Lite
Triton models (.trt, .engine)
OpenVino IR Models

Microcontroller based Computer Vision (Embedded Computer Vision)

Power Ranges: 3.3 - 9V DC, 0.1 - 2 A

Programming Languages : Micropython , C++ , Assembly C

Libraries: TFLite, TinyML- Formats any, Open Vino (Still under experimentation)

Frameworks: Linux, Assembly Level, RISC-OS (V)

What are these devices for?

These are basically very low powered devices used for mostly Single task based Computer Vision in a more efficient manner and consuming as less current as possible.

Sipeed MAIX Go Suit
Open MV M7 —
Open MV H7 Plus —-
Open CV AI Kit
Tiny Vision based FPGA Board (FPGA - Field Programmable Gate Arrays) etc.

These devices are less costly and low maintenance too as they don’t need extra efforts to cool the systems and also to power the systems.