- Ubuntu 18.04
- Python 3.7
- OpenVINO 2020.2.120
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
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
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
For video :
-
Play
andPause
-
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)
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
Currently you can only select 1 target hardware
Only FP16 and FP32
Please see this
- 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)
- Supports most of Object Detection models from Open Model Zoo
- Supports USB Webcam and Youtube Video as video source
- Added Yolo support
- Added tooltips to UI components
- Bug fixes
- Added Async Inference
- Added Open Pose model