Skip to content

Latest commit

 

History

History
131 lines (82 loc) · 2.55 KB

README.md

File metadata and controls

131 lines (82 loc) · 2.55 KB
Raw

Object Detection app

Environment

  • Ubuntu 18.04
  • Python 3.7
  • OpenVINO 2020.2.120

Ubuntu Setup

OpenVINO toolkit must be installed on the device.

Follow this instruction to setup Ubuntu System

Follow this instruction to setup Raspberry Pi

Important

Raspberry Pi requires Movidius to run inference

App Setup

Clone this repo and run /App/ObjectDetectino/Python/setup.sh

cd ~ && \
git clone https://github.com/daisukeiot/OpenVINO-Toolkit-Setup.git && \
cd OpenVINO-Toolkit-Setup/App/ObjectDetection/Python && \
./setup.sh

Running App

Start the app with :

cd ~/OpenVINO-Toolkit-Setup/App/ObjectDetection/Python
python3.7 ./main.py
  • Access the UI from a browser http://<IP Address of the device>:8080

Browser

Video Playback

For video :

  • Play and Pause

  • Playback Mode

    • Sync Playback Mode

      Plays at the frame rate of the source video

    • Performance Mode

      Plays video as fast as the system can

For camera :

  • Change resolution
    Currently the app supports following resolutions (*camera must support selected resolution)

    • VGA (640x480)
    • XGA (1024x748)
    • HD (1280x720)
    • WXGA (1280x800)
    • FHD (1920x1080)

AI Inference

You can start and stop AI Inference.

Select Model, Hardware, and Precision to run inference on then click "Configure" button

  • Use toggle switch to turn on/off AI inference
  • Look for the model locally, if not found, download from internet
  • Converts model to IR if necessary
  • Loads model to selected HW

Target HW

Currently you can only select 1 target hardware

Precision

Only FP16 and FP32

Run in container

Please see this

To do list

  • Add Yolo support
    Done : 4.23.2020 (ver 1.1)
  • Add tooltips to buttons Done : 4.27.2020 (ver 1.2)
  • Add Async Inference
  • Containerization
  • Add IoT Hub device app/IoT Edge support
  • HETERO support
  • Improve status line display/logging Done : 4.27.2020 (ver 1.2)
  • Handle camera error Done : 4.27.2020 (ver 1.2)

Release Notes

Version 1 (April 21, 2020)

  • Supports most of Object Detection models from Open Model Zoo
  • Supports USB Webcam and Youtube Video as video source

Version 1.1 (April 23, 2020)

  • Added Yolo support

Version 1.2 (April 27, 2020)

  • Added tooltips to UI components
  • Bug fixes

Version 1.3 (May 7, 2020)

  • Added Async Inference

Version 1.4 (May 27, 2020)

  • Added Open Pose model