Genio 510 QSG

Genio 510 EVK Demo Requirements

This demonstration kit is very flexible and can be setup in various configurations, with software support for Yocto or Android. However for the purpose of this Quick Start Guide, a specific configuration is chosen with the following conditions:

  • IOT-YOCTO software build

  • D6 Camera Board enabled

Additional items needed for Quick Start Guide demo

To support the demo, the following additional items are needed.

  • Two USB Cables (USB-2.0-A to Micro-B)

  • External PC for loading flash and interacting via command prompt

Setup PC Environment

  • Install the Genio Tools on your Linux or Windows PC.

  • Setup serial port for viewing debug output and interacting with the kit through a command line interface

  • Check Device Manager that the USB COM port appears as a UART device under COM Ports

  • Install a terminal program as necessary, for example Putty. Set baud rate speed to 921600 and open the corresponding port to connect UART0.

../_images/Example-on-Putty.png

Example on Putty

Kit Hardware Installation

  • Insert the D6 Camera Board to CS10 on the main motherboard.

../_images/hw_evk_g510-evk_with_D6.jpg

  • Change jumper settings as required in Yocto operation.

Jumper

Position

J1

2-3

J2

2-3

J5

2-3
../_images/d6_camera_board_jumpers.jpg

Camera Board Installation

  • Connect a USB cable to connector labeled UART0 as shown in the image USB Port Locations below:

This requires installation of the Google Android USB driver as outlined in the Setup PC Environment section. If using Windows, open Device Manager and confirm that the driver appears under Android Device \ Android Bootloader Interface.

  • Connect a second USB cable to USB0 as shown in the image USB Port Locations below:

This requires installation of the FTDI VCP USB driver as outlined in the Setup PC Environment section. If using Windows, open Device Manager and confirm that the driver appears under Ports (COM & LPT) \ USB Serial Port (COMx), where x is the port assigned by your system.

../_images/USB_ports.png

USB Port Locations

Further details regarding both USB drivers can be found in the flash trouble-shooting section.

  • Plug AC/DC power adapter into the connector.

Flash the software

This guide shows how to “flash” (write) the demo image into the non-volatile flash memory of the main motherboard. The flashing process is performed through the USB0 connection. In order to write the on-board storage with Genio Tools, you need to set the IoT SoC in download mode, which allows Genio Tools to transfer a download agent binary to the SRAM of SoC. The download agent then provides a fastboot interface for subsequent image transfer and storage write operations.

Note

Avoid having a long path name by installing to a root folder and shortening the file name if necessary, for example C:\\evk.

  • Open a Windows cmd window (to enter commands)

  • Change working folder to the image directory before running the genio-flash commands:

cd C:\evk

  • Flash the image using the overlay which enables both cameras. which enables Before flashing the board, check if there are certain board features you’d like to enable:

genio-flash --load-dtbo camera-ar0830-ap1302-csi0.dtbo --load-dtbo gpu-mali.dtbo --load-dtbo apusys.dtbo --load-dtbo video.dtbo

  • Once you see the Looking for a MediaTek SoC ... prompt, start the process to boot the board in flash download mode.

  1. Press and keep pressing the Download button.

  2. Press and release the RST button

  3. Hold the Download button until you see the log Erasing 'mmc0' message, as shown below:

../_images/g700_download_buttons.gif

Reset the board to enter download mode

You should see flashing process started after releasing the volume up button. A typical successful log looks something like this:

Genio Tools: v1.4
Yocto Image:
        name:     Rity Demo Image (rity-demo-image)
        distro:   Rity Demo Layer 22.3-dev (rity-demo)
        codename: dunfell
        machine:  genio-510-evk
        overlays: []

Looking for MediaTek SoC matching USB device 0e8d:0003
Opening COM11 using baudate=115200
Connected to MediaTek MT8370 SoC
Sending bootstrap to address: 0x201000
Jumping to bootstrap at address 0x201000 in AArch64 mode
erasing mmc0
< waiting for any device >
Erasing 'mmc0'                                     (bootloader) request sz: 0x3a6800000, real erase len: 0x0
OKAY [  0.233s]

… (shortened for brevity)

Rebooting                                          OKAY [  0.002s]
Finished. Total time: 0.003s

Note

The above command should be the only needed to flash the image with the configurations necessary in this QSG. Further background and troubleshooting information including different command options are available in Flash Genio 510-EVK. A common problem is that the download does not start and simply keeps waiting for the device. More information is available in Missing “Yocto” Device Driver (Windows Only).

  • Once the kit boots up with the new flash image, debug log messages can be seen through UART0 in the Putty terminal window.

../_images/Log-Setting.png

View of debug log messages appearing from UART0 upon board bootup

  • Verify the desktop boot screen appears on the display of the kit.

../_images/yocto_boot_screen_low_res1.jpg

  • Log into board through UART0 interface.

Once board has fully booted, a login prompt is sent across UART0 similar to that shown below.

genio-510-evk login:

Type root, then enter, to log in.

NNStreamer Demos

NNStreamer is a set of GStreamer plugins that allow GStreamer developers to adopt neural network models, and neural network developers to manage neural network pipelines with their filters in a easy and efficient way. In order to demonstrate these features we first set up the camera with the v4l2-ctl command.

Camera Setup

We must first find the V4L2 device for the YUV sensor. There is only one such device which is for the preview stream. The user can list all devices by typing the following command and then searching for the device named mtk-v4l2-camera:

v4l2-ctl --list-device
    MediaTek MDP3 ();
            /dev/video2

… (shortened for brevity)

mtk-v4l2-camera (platform:mtkcam0):
        /dev/video5

Important

The above results are shown only as an example. The node number may change every time the system boots. In this case the mtk-v4l2-camera device is at /dev/video5. This path and node number 5 will be used in the demos below.

Before using the cameras, you must set the resolution and the pixel format to 1920 * 1080, UYUV using the v4l2-ctl command. The command requires several parameters, one of which is the media device of the sensor. The index of the media device in this example is /dev/video5 which was determined by the results above.

v4l2-ctl -d /dev/video5 --set-fmt-video=width=1920,height=1080,pixelformat=UYVY --stream-mmap --stream-count=10 --stream-to=/tmp/ap1302.yuv --verbose

Use gst-launch-1.0 to launch the camera.

gst-launch-1.0 v4l2src device=/dev/video5 ! video/x-raw,width=1920,height=1080,format=UYVY ! v4l2convert output-io-mode=dmabuf-import ! video/x-raw,width=1280,height=720 ! waylandsink sync=false
../_images/sw_yocto_app-dev_camera_yuv-show-on-weston.png

Show Preview stream on Weston

Confirm that a live video stream is being shown on the LCM. Before proceeding to the next demos, type CTRL+C to end this video stream.

CTRL+C

Further details about hardware operation, background theory, or troubleshooting can be found in the BSP User Guide - Launch Camera section.

NNStreamer Environment Setup

The following NNStreamer demos are run using a python script run_nnstreamer_example.py. This script is configured using environment variables specifying which camera type to use, which mode to use, which processing engine to use, and the device node of your camera as determined in the step above. Set the following variables before running any of the following demos.

CAM_TYPE=yuvsensor
MODE=1
ENGINE=neuronsdk
CAMERA_NODE_ID=5

Important

The node number may change every time the system boots. As determined above, the mtk-v4l2-camera device in this example is at /dev/video5. This node number 5 is used for the CAMERA_NODE_ID variable. Further details can be found in the page NNStreamer.

Pose Detection

Set the APP variable so the script runs the proper demo. Then run the script.

APP=pose_estimation
python3 /usr/bin/nnstreamer-demo/run_nnstreamer_example.py --app $APP --cam_type $CAM_TYPE --cam $CAMERA_NODE_ID --engine $ENGINE --performance $MODE

A video stream from the camera should now be displayed on the LCM however with indicators similar to the below image.

../_images/tools_nnstreamer_examples_pose_estimation.png

Before proceeding to the next demos, type CTRL+C to end this video stream.

CTRL+C

Face Detection

Set the APP variable so the script runs the proper demo. Then run the script.

APP=face_detection
python3 /usr/bin/nnstreamer-demo/run_nnstreamer_example.py --app $APP --cam_type $CAM_TYPE --cam $CAMERA_NODE_ID --engine $ENGINE --performance $MODE

A video stream from the camera should now be displayed on the LCM however with the face blocked out as shown in the below image.

../_images/tools_nnstreamer_examples_face_detection.png

Before proceeding to the next demos, type CTRL+C to end this video stream.

CTRL+C

OpenGL ES Demos

On IoT Yocto, there are several OpenGL ES applications that can used to check that the GPU is running and well integrated within the Wayland/Weston system. In this section we introduce two GLES applications: weston-simple-egl and kmscube. They both use the GPU to render content, but they have different approaches to creating their graphic windows.

Display a rotating cube

This is a simple GLES program demonstrating how to drive bare metal graphics without a compositor (Wayland), using DRM/KMS (kernel mode setting), GBM (graphics buffer manager), and EGL for rendering content using OpenGL ES.

After having typed CTRL+C from the above step, you should see only the simple desktop window which shows the date and time in the upper right hand corner. This desktop graphic is being driven by the Weston service. In this demo, we will display a rotating cube by running the kmscube command. However kmscube will create its own window on which it renders content. So, before we launch kmscube, we must make sure that no other window system occupies the display. This means stopping the Weston process that drives the desktop.

  • Stop weston service if it already started.

    systemctl stop weston

  • Confirm that the desktop disappears and you see only a dark screen on the HDMI monitor.

  • Launch kmscube and confirm that you can see a rotating cube as shown in below figure.

    kmscube
../_images/sw_yocto_gpu-kmscube-screenshot.png

Screenshot of kmscube

  • Type CTRL+C to stop displaying the rotating cube.

Display a rotating triangle

This is a simple demonstration GLES program that draws a rotating colorized triangle on a transparent Wayland window. This program is not responsible for creating the window, and instead merely renders its content on an existing window which Weston creates. So, before launching it, we must start Weston.

  • Start weston service if it is not started yet.

    systemctl start weston

  • Confirm that the desktop reappears which shows the date and time in the upper right hand corner.

  • Launch weston-simple-egl and confirm that you can see a rotating triangle as shown in below figure.

    weston-simple-egl
153 frames in 5 seconds: 30.600000 fps
150 frames in 5 seconds: 30.000000 fps
...

  • Type CTRL+C to stop displaying the rotating triangle.

../_images/sw_yocto_gpu-weston-simple-egl-screenshot.png

Screenshot of weston-simple-egl

Additional Details

You can find source code for weston-simple-egl in the Weston codebase: $BUILD_DIR/tmp/work/aarch64-poky-linux/weston/8.0.0-r0/weston-8.0.0/clients/simple-egl.c or weston github

You can find source code in Weston codebase: $BUILD_DIR/tmp/work/aarch64-poky-linux/kmscube/git-r0/git or mesa gitlab

Further details about the GPU, background theory, or troubleshooting can be found in the BSP User Guide - GPU section.