Difference between revisions of "Template:RK3399-Android8"

From FriendlyELEC WiKi
Jump to: navigation, search
(updated by API)
(updated by API)
Line 20: Line 20:
 
*[[FriendlyThings for RK3399|Android: FriendlyThings for RK3399]]
 
*[[FriendlyThings for RK3399|Android: FriendlyThings for RK3399]]
 
*[[FriendlyThings APIs|Android: FriendlyThings APIs]]
 
*[[FriendlyThings APIs|Android: FriendlyThings APIs]]
 +
 +
===Android Neural Networks Sample===
 +
Android8.1 BSP预装了Rockchip为Android8.1提供的Android NN SDK,支持GPU加速。<br />
 +
我们提供的Android8固件中预装了Android NN的示例程序:TfLiteCameraDemo,这是一个使用Rockchip AndroidNN GPU加速的MobileNet分类器Demo,运行时需要主板连接有摄像头模块,USB摄像头和CSI摄像头都是支持的,示例程序通过点击首页的 “TfLiteCameraDemo”图标进入,运行效果如下所示:<br /><br />
 +
[[File:TfLiteCameraDemo.png|frameless|600px]]<br />
 +
关于Android NN SDK的详细描述,可以参考原厂的这份文档:[https://gitlab.com/friendlyelec/rk3399-android-8.1/blob/master/RKDocs/rk3399/RK3399_Android8.1_AndroidNN_SDK_V1.0_20180605%E5%8F%91%E5%B8%83%E8%AF%B4%E6%98%8E.pdf [RK3399_Android8.1_AndroidNN_SDK_V1.0_20180605发布说明.pdf]]<br />
 +
TfLiteCameraDemo示例源代码下载地址:[https://github.com/rockchip-linux/tensorflow/tree/master/tensorflow/contrib/lite/java/demo [TfLiteCameraDemo源代码]]<br />

Revision as of 10:14, 13 November 2018

1 Work with Android 8.1

FriendlyElec provides a full Android8.1 BSP for {{{1}}}. The source code is hosted at gitlab.com and is open source. The BSP supports GPU and VPU hardware acceleration.

1.1 Connect MIPI Camera to T4

FriendlyElec developed a MIPI camera CAM1320 for {{{1}}} and it works under Android. You can use this camera to take pictures and record video. The operation is straightforward. You just need to connect the camera at your {{{1}}}'s MIPI interface, boot your T4 to Android and start Android's camera app.

The {{{1}}} has two MIPI interfaces: one front camera and the other rear camera:

Interface Camera Type(front or rear)
MIPI-CSI1 Android rear camera
MIPI-CSI2 Android front camera

1.2 Android8.1 hardware access

You can use the FriendlyThings SDK to access and control hardware resources on the motherboard in the Android App, such as Uart, SPI, I2C, GPIO and other interfaces. For details, please refer to the following two documents.:

1.3 Android Neural Networks Sample

Android8.1 BSP预装了Rockchip为Android8.1提供的Android NN SDK,支持GPU加速。
我们提供的Android8固件中预装了Android NN的示例程序:TfLiteCameraDemo,这是一个使用Rockchip AndroidNN GPU加速的MobileNet分类器Demo,运行时需要主板连接有摄像头模块,USB摄像头和CSI摄像头都是支持的,示例程序通过点击首页的 “TfLiteCameraDemo”图标进入,运行效果如下所示:

TfLiteCameraDemo.png
关于Android NN SDK的详细描述,可以参考原厂的这份文档:[RK3399_Android8.1_AndroidNN_SDK_V1.0_20180605发布说明.pdf]
TfLiteCameraDemo示例源代码下载地址:[TfLiteCameraDemo源代码]