yolov8

YOLOV8 example

This package provides 3 yolov8 sample that running on RyzenAI and OnnxRuntime(VitisAI EP) . It allows user to directly run yolov8 on a IPU laptop. Please refer to the Ryzen AI SW platform installation guide to install all the dependent packages.

How to build:

1. move yolov8(path: example/yolov8) folder into the 'yolov8_e2e' folder (path: tutorial/yolov8_e2e/..)

└── yolov8_e2e
    ├── bin
    ├── code 
    ├── images
    ├── yolov8
           └── yolov8
    └── ... 

2. build with batch file

cd tutorial\yolov8_e2e\yolov8
.\build.bat

Run

Jpeg sample

.\run_jpeg.bat .\DetectionModel_int.onnx .\sample_yolov8.jpg

Output:
result: 0 person 490.38498 85.79535 640.00488 475.18262 0.932453
result: 0 person 65.96048 97.76373 320.66068 473.83783 0.924142
result: 0 person 182.15485 306.91266 445.14795 475.26132 0.893309
result: 27 tie 584.48022 221.15732 632.27008 244.21243 0.851953
result: 27 tie 175.62622 224.15210 235.84900 248.83557 0.651355

Simple camera sample

.\camera.bat .\DetectionModel_int.onnx

Press esc to exit

Live camera sample

.\camera_nx.bat

Press esc to exit

Environment and arguments

set NUM_OF_DPU_RUNNERS=4 to use 4x1x4

.\camera_nx.bat -h

C:\Users\ibane\Desktop\voe-win_amd64-with_xcompiler_on-c07e419-latest\yolov8\..\bin\camera_yolov8_nx1x4 [options...]
Options:
      -c [parallel runs]: Specifies the (max) number of runs to invoke simultaneously. Default:1.
      -s [input_stream] set input stream, E.g. set 0 to use default camera.
      -x [intra_op_num_threads]: Sets the number of threads used to parallelize the execution within nodes, A value of 0 means ORT will pick a default. Must >=0.
      -y [inter_op_num_threads]: Sets the number of threads used to parallelize the execution of the graph (across nodes), A value of 0 means ORT will pick a default. Must >=0.    
      -D [Disable thread spinning]: disable spinning entirely for thread owned by onnxruntime intra-op thread pool.
      -Z [Force thread to stop spinning between runs]: disallow thread from spinning during runs to reduce cpu usage.
      -T [Set intra op thread affinities]: Specify intra op thread affinity string.
         [Example]: -T 1,2;3,4;5,6 or -T 1-2;3-4;5-6
         Use semicolon to separate configuration between threads.
         E.g. 1,2;3,4;5,6 specifies affinities for three threads, the first thread will be attached to the first and second logical processor.
      -R [Set camera resolution]: Specify the camera resolution by string.
         [Example]: -R 1280x720
         Default:1920x1080.
      -r [Set Display resolution]: Specify the display resolution by string.
         [Example]: -r 1280x720
         Default:1920x1080.
      -L Print detection log when turning on.
      -h: help

1. -c [set channels of yolov8 session]. Recommended as 5-6(Can be real-time and low latency).
2. -x & -y [set num threads of inter and intra op]. Both recommended as 1. Higher values in this(or most) cases do not result in any performance gain but lead to a significant increase in CPU power consumption.
3. -D [Disable thread spinning]. Enable spinning-wait by default, it will causes CPU usage to 100%. We recommended adding '-D', which significantly reduces power with impact on performance
4. -R/-r Please set it according to your computer configuration. Default:[FHD/FHD]
5. -L Printing the log may reduce the frame rate, generally recommended to turn off.
6. -s stream-id, default camera is usually 0.

Usage

1. Recommended(Generally can run in real time(30fps))

%cd%\..\bin\camera_yolov8_nx1x4  -c 5 -x 1 -y 1 -s 0 -D 

2. More target may be available

%cd%\..\bin\camera_yolov8_nx1x4  -c 7 -x 1 -y 1 -s 0 -D 

3. Lower camera resolution and higher display resolution

%cd%\..\bin\camera_yolov8_nx1x4  -c 5 -x 1 -y 1 -s 0 -D -R 1280x720 -r 2560x1440

4. With video file

%cd%\..\bin\camera_yolov8_nx1x4  -c 5 -x 1 -y 1 -s video/test.mp4 -D 

Sample Code

  auto model = Yolov8Onnx::create(std::string(argv[1]), 0.3); // Create session

  cv::Mat image = cv::imread(argv[2]); // Load a image

  auto results = model->run(image); // Run

  process_result(image, results); // Draw box

Other recommendations

1. About Power mode: Windows has 3 power by default. they are Best power efficiency, Balenced and Best Performance. They will have different performance scheduling strategies and power consuption performance. Power plug-in and battery are also difference. A brief comparison of them is as follows(tested Live camera sample on HP 835g10):

Best power efficiency with plug-in:  fps≈30 | power≈11w
Balenced with plug-in:  fps≈30 | power≈15w
Best Performance with plug-in: fps≈30 | power≈15w
Best power efficiency with battery: fps≈30 | power≈10w
Balenced with plug-in and without -D, -x 1, -y 1: fps≈30 | power≈26w

To configure power mode in Windows 11 by following steps: Open Settings by pressing Win + I or right-clicking the Start menu and selecting Settings.

Go to System > Power (or Power and Battery on laptops).
Find the Power mode drop-down list and select one of the following options.

2. Display setting To make the display magnification of opencv normal, please temporarily set the display scaling rate to 100%. To adjust display scaling in Windows 11, Please: Open Settings.

Click on System.
Click on Display.
Under the "Scale and layout" section, use the drop-down menu and select.