IMPORTANT: This page is being updated and the content might not be 100% correct.
Engine: Movidius
Movidius neural compute stick (NCS) is an USB device containing Myriad 2, the DLA (Deep Learning Accelerator) chip.

| Feature | Description |
|---|---|
| Inference Performance | FPS 7.5 for object detection (MobileNet SSD) |
| Hardware Performance | 80~150 GFLOPS [1] |
| Power Consumption | 1W power envelope [2] |
BerryNet Movidius Engine helps user to create edge AIoT applications in minutes, and is powered by OpenVINO.
| RPi3 w/ 2-NCS | UP Squared (w/ Myriad 2) |
|---|---|
![]() |
![]() |
Enable Movidius Engine
Here are the steps to enable the support:
-
Install dependencies and inference engine:
$ bash utils/install-movidius.sh -
Update BerryNet inference engine settings
- Edit
/usr/local/bin/berrynet-manager- Replace
detection_fast_server.servicebyclassify_movidius_server.service
- Replace
- Edit
/usr/local/berrynet/config.js- Set
config.inferenceEnginefromdetectortoclassifier
- Set
- Edit

