借助AI Express实现手势识别

asdf2020-08-10

2121

一、背景

手势识别是互动娱乐、智能车载等领域中的常用功能之一。AI-Express 2.4.0版本中，结合地平线基于序列的行为识别模型，给出了手势识别参考方案。

本教程中您将了解到如何在AI-Express中运行模型和算法策略，构建Workflow，并搭建智能应用。

二、方案介绍

手势识别solution的整体流程如下。它包含输入，手检测，跟踪，手关键点检测，手势识别，投票，输出这几大部分。

手检测运行Faster R-CNN模型，以检测框的方式给出手在画面中的位置。

之后会对手进行跟踪。在这一步中，检测框将被分配track_id，它们的坐标也会根据历史轨迹被修正。

手部关键点检测运行CNN模型，基于输入的图片和检测框对手部关键点进行预测。

手势识别运行CNN模型，它基于手部关键点序列给出手势预测结果。目前支持的手势共14种：无手势，单个手指指，两个手指指，单手指点击，单手指双击，两个手指点击，两个手指双击，向上抛掷，向下抛掷，向左抛掷，向右抛掷，放大，缩小，开合。

为了得到更平滑的检测结果，我们在手势识别模块后添加了投票模块。利用滑窗，结合多帧上的预测给出最终结果。

三、教程

0 Prerequisite

您需要了解地平线AI-Express应用开发中间件，熟悉Linux平台下的代码编译、部署。

1 开发环境（软硬件、编译环境）

本教程基于AI Express软件包和地平线96Board AI开发板/X3-Dev开发板。您可以参考软件包文档中的“快速上手”章节，准备开发环境。

2 开发教程

本教程将首先介绍如何通过XStream Method使用模型。之后说明如何将不同Method串联起来组成workflow。最后简述如何利用XProto搭建具有手势识别功能的智能应用。

2.1 使用XStream Method

如方案介绍中所述，一个完整流程中包含许多小而具体的任务。AI Express通过不同的Method承载不同的具体任务。一个Method或运行算法模型，或执行算法策略。在手势识别solution使用了FasterRCNNMehtod，CNNMethod，MotMethod，和VoteMethod。前两者运行模型，后两者执行算法策咯。该solution中没有新建Method，复用现有Method的方式十分简单。您只需要通过Method配置文件指定模型文件和其他相关信息即可（包括但不限于：模型输出，后处理参数）。

FasterRCNNMethod挂载手检测模型。solution中使用的是一个多任务模型，实际上我们只需关注手检测结果。Method配置文件如下：

手部关键点检测模型和手势识别模型都挂载在CNNMethod，通过不同配置文件执行不同的运算。Method配置文件如下：

MotMethod和VoteMethod分别承载手跟踪和识别结果投票任务。配置文件相对简单易懂，不在此列出。

2.2 搭建workflow

上文介绍了使用XStream Method运行模型或算法策略。想要实现完整的算法功能，您需要构建Workflow，串联各Method，实现模块间的数据互通。

目前，XStream Workflow通过Json文件表示。一个简单的Workflow框架如下所示。Workflow中包含input, output, workflow三大部分，分别代表Workflow的输入，Workflow的输出，和Workflow中包含的Mehtods。“workflow”是一个Json数组，其中每一个Json对象代表一个Method。Json对象中的各键值的意义可以参考下面的注释。

在Workflow中串联各Method，就是基于拓扑序连接它们的输入输出。完整的手势识别Workflow如下所示。需要注意Method输入输出的名字不能重复且要严格对应。

2.3 利用XProto搭建智能应用

至此，您已经了解了如何实现手势识别solution的智能部分。除了智能部分，您还需要获取输入数据，解析并发送智能数据。您可以利用XProto搭建一个完整的智能应用。

本方案使用到了XProto中的VioPlugin, SmartPlugin, 和WebSocketPlugin。

VioPlugin：VioPlugin负责获取、转换图像数据并控制图像数据获取速率，然后将图像数据或丢帧消息推送给消息总线。

SmartPlugin：SmartPlugin是基于XStream通用SDK接口开发的通用智能应用运行模块。它监听XProto总线上面的VioPlugin送来的“XPLUGIN_IMAGE_MESSAGE”类型消息，并且将它送到XStream的Workflow实例里面，同时接收Workflow输出的结果，并将其序列化后产生智能消息然后送回到消息总线上。

WebSocketPlugin：WebSocketPlugin监听总线上SmartPlugin发来的消息。如果有web端发来的请求，WebSocketPlugin在处理请求后会主动发送视频数据和智能数据。智能数据基于Proto Buffer进行封装。

您需要在SmartPlugin中，根据您在workflow中指定的输出名称，以对应类型解析智能结果。手势识别solution中，识别结果以Attribute类型输出。VioPlugin和WebSocketPlugin不需要做改动。然后，只需要在主函数中进行各plugin的初始化和启动等操作即可。

2.4 编译运行

您可以根据所使用的硬件在ai-express-release包中运行“bash build.sh [x2|x3] && bash deploy.sh”进行编译和打包，命令成功执行后会生成deploy部署包。

您可以通过deploy包内的run.sh脚本运行智能应用。应用运行后，您可以在命令行看到相关智能结果的输出。AI-Express 2.2.0及以上版本包含web展示端，您也可以访问板子IP，通过web展示端查更直观地查看应用运行结果。

感知

0/1000

张建宏
Lv.1
模型打包成 deploy时出现一些错误
[zjh@localhost AI-EXPRESS]$ bash deploy.sh
+ ALL_PROJECT_DIR=/opt/AI-EXPRESS
+ RELEASE_DIR=/opt/AI-EXPRESS/deploy
+ rm /opt/AI-EXPRESS/deploy -rf
+ mkdir -p /opt/AI-EXPRESS/deploy
++ cat platform.tmp
+ ARCHITECTURE=x3
+ mkdir /opt/AI-EXPRESS/deploy/lib/
+ cp /opt/AI-EXPRESS/build/lib/libvioplugin.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/build/lib/libsmartplugin.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/build/lib/libvisualplugin.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/build/lib/libwebsocketplugin.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ '[' x3 == x3 ']'
+ cp /opt/AI-EXPRESS/build/lib/libuvcplugin.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp '/opt/AI-EXPRESS/deps/x3_prebuilt/lib/libguvc*.so' /opt/AI-EXPRESS/deploy/lib/ -rf
cp: 无法获取'/opt/AI-EXPRESS/deps/x3_prebuilt/lib/libguvc*.so' 的文件状态(stat): No such file or directory
+ cp /opt/AI-EXPRESS/build/lib/libxstream-media_codec.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp '/opt/AI-EXPRESS/build/lib/libmulti*.so' /opt/AI-EXPRESS/deploy/lib/ -rf
cp: 无法获取'/opt/AI-EXPRESS/build/lib/libmulti*.so' 的文件状态(stat): No such file or directory
+ cp '/opt/AI-EXPRESS/build/lib/libgdcplugin*.so' /opt/AI-EXPRESS/deploy/lib/ -rf
cp: 无法获取'/opt/AI-EXPRESS/build/lib/libgdcplugin*.so' 的文件状态(stat): No such file or directory
+ cp '/opt/AI-EXPRESS/source/solution_zoo/apa/gdcplugin/deps/*.so' /opt/AI-EXPRESS/deploy/lib/ -rf
cp: 无法获取'/opt/AI-EXPRESS/source/solution_zoo/apa/gdcplugin/deps/*.so' 的文件状态(stat): No such file or directory
+ cp '/opt/AI-EXPRESS/build/lib/libdisplayplugin*.so' /opt/AI-EXPRESS/deploy/lib/ -rf
cp: 无法获取'/opt/AI-EXPRESS/build/lib/libdisplayplugin*.so' 的文件状态(stat): No such file or directory
+ cp '/opt/AI-EXPRESS/source/solution_zoo/apa/displayplugin/deps/lib/*.so' /opt/AI-EXPRESS/deploy/lib/ -rf
cp: 无法获取'/opt/AI-EXPRESS/source/solution_zoo/apa/displayplugin/deps/lib/*.so' 的文件状态(stat): No such file or directory
+ cp /opt/AI-EXPRESS/build/lib/libcanplugin.so /opt/AI-EXPRESS/deploy/lib/ -rf
cp: 无法获取'/opt/AI-EXPRESS/build/lib/libcanplugin.so' 的文件状态(stat): No such file or directory
+ cp /opt/AI-EXPRESS/build/lib/libanalysisplugin.so /opt/AI-EXPRESS/deploy/lib/ -rf
cp: 无法获取'/opt/AI-EXPRESS/build/lib/libanalysisplugin.so' 的文件状态(stat): No such file or directory
+ cp /opt/AI-EXPRESS/deps/bpu_predict/x3/lib/libbpu_predict.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/x3_prebuilt/lib/libhbrt_bernoulli_aarch64.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/protobuf/lib/libprotobuf.so.10 /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/opencv/lib/libopencv_world.so.3.4 /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/live555/lib/libBasicUsageEnvironment.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/live555/lib/libgroupsock.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/live555/lib/libliveMedia.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/live555/lib/libUsageEnvironment.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/uWS/lib64/libuWS.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/xwarehouse/lib/libxwarehouse.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/zeroMQ/lib/libzmq.so /opt/AI-EXPRESS/deps/zeroMQ/lib/libzmq.so.5 /opt/AI-EXPRESS/deps/zeroMQ/lib/libzmq.so.5.2.0 /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/source/common/xstream/python_api/package/lib/libmethod_factory.so /opt/AI-EXPRESS/source/common/xstream/python_api/package/lib/vision_type.so /opt/AI-EXPRESS/source/common/xstream/python_api/package/lib/xstream_internal.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/source/common/xproto/python_api/package/lib/native_xproto.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/libyuv/lib/libyuv.so /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/deps/libjpeg-turbo/lib/libturbojpeg.so /opt/AI-EXPRESS/deps/libjpeg-turbo/lib/libturbojpeg.so.0 /opt/AI-EXPRESS/deps/libjpeg-turbo/lib/libturbojpeg.so.0.1.0 /opt/AI-EXPRESS/deploy/lib/ -rf
+ cp /opt/AI-EXPRESS/run.sh /opt/AI-EXPRESS/deploy/ -rf
+ cp /opt/AI-EXPRESS/start_nginx.sh /opt/AI-EXPRESS/deploy/ -rf
+ mkdir /opt/AI-EXPRESS/deploy/configs/
+ '[' x3 == x2 ']'
+ '[' x3 == x3 ']'
+ cp /opt/AI-EXPRESS/output/vioplugin/config/vio /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.2610 /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.2610.hg /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.96board /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.96board.hg /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.j3dev.cam /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.j3dev.fb /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.j3dev.multi_cam_async /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.j3dev.multi_cam_sync /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.j3dev.multi_fb_sync /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.x3dev.cam /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.x3dev.fb /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.x3dev.fb_cam /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.x3dev.multi_cam_async /opt/AI-EXPRESS/output/vioplugin/config/vio_config.json.x3dev.multi_cam_sync /opt/AI-EXPRESS/output/vioplugin/config/vio_hg /opt/AI-EXPRESS/deploy/configs/ -rf
+ echo 'copy viowrapper configs'
copy viowrapper configs
+ cp /opt/AI-EXPRESS/source/common/viowrapper/config/x3dev/hb_camera_x3.json /opt/AI-EXPRESS/source/common/viowrapper/config/x3dev/hb_vio_x3_1080_fb.json /opt/AI-EXPRESS/source/common/viowrapper/config/x3dev/hb_vio_x3_1080.json /opt/AI-EXPRESS/source/common/viowrapper/config/x3dev/hb_vio_x3_608.json /opt/AI-EXPRESS/deploy/configs/ -rf
+ cp /opt/AI-EXPRESS/output/apa /opt/AI-EXPRESS/deploy/ -rf
cp: 无法获取'/opt/AI-EXPRESS/output/apa' 的文件状态(stat): No such file or directory
+ cp '/opt/AI-EXPRESS/output/apa/configs/configs/*' /opt/AI-EXPRESS/deploy/configs/ -rf
cp: 无法获取'/opt/AI-EXPRESS/output/apa/configs/configs/*' 的文件状态(stat): No such file or directory
+ cp /opt/AI-EXPRESS/output/multivioplugin/bin/multivioplugin_test /opt/AI-EXPRESS/deploy/apa/ -rf
cp: 无法获取'/opt/AI-EXPRESS/output/multivioplugin/bin/multivioplugin_test' 的文件状态(stat): No such file or directory
+ cp /opt/AI-EXPRESS/output/multisourceinput /opt/AI-EXPRESS/deploy/ -rf
cp: 无法获取'/opt/AI-EXPRESS/output/multisourceinput' 的文件状态(stat): No such file or directory
+ cp /opt/AI-EXPRESS/output/visualplugin/config/visualplugin_body.json /opt/AI-EXPRESS/output/visualplugin/config/visualplugin_face.json /opt/AI-EXPRESS/output/visualplugin/config/visualplugin_vehicle.json /opt/AI-EXPRESS/deploy/configs/ -rf
+ cp /opt/AI-EXPRESS/source/common/xproto/plugins/websocketplugin/configs/websocketplugin_attribute.json /opt/AI-EXPRESS/deploy/configs/ -rf
+ mkdir -p /opt/AI-EXPRESS/deploy/models
+ cp /opt/AI-EXPRESS/models/x3/faceAgeGender/so/faceAgeGender.hbm /opt/AI-EXPRESS/models/x3/faceMask/so/faceMask.hbm /opt/AI-EXPRESS/models/x3/faceMultitask/so/faceMultitask.hbm /opt/AI-EXPRESS/models/x3/faceQuality/so/faceQuality.hbm /opt/AI-EXPRESS/models/x3/personMultitask/so/personMultitask.hbm /opt/AI-EXPRESS/deploy/models/ -rf
+ '[' x3 == x3 ']'
+ cp /opt/AI-EXPRESS/models/x3/SegmentationMultitask_1024x768/so/SegmentationMultitask_1024x768.hbm /opt/AI-EXPRESS/deploy/models/ -rf
cp: 无法获取'/opt/AI-EXPRESS/models/x3/SegmentationMultitask_1024x768/so/SegmentationMultitask_1024x768.hbm' 的文件状态(stat): No such file or directory
+ cp /opt/AI-EXPRESS/models/x3/personMultitask_1024x768/so/personMultitask_1024x768.hbm /opt/AI-EXPRESS/deploy/models/personMultitask_1024x768.hbm -rf
cp: 无法获取'/opt/AI-EXPRESS/models/x3/personMultitask_1024x768/so/personMultitask_1024x768.hbm' 的文件状态(stat): No such file or directory
+ cp '/opt/AI-EXPRESS/source/solution_zoo/apa/models/*' /opt/AI-EXPRESS/deploy/models/ -rf
cp: 无法获取'/opt/AI-EXPRESS/source/solution_zoo/apa/models/*' 的文件状态(stat): No such file or directory
+ cp /opt/AI-EXPRESS/output/face_solution /opt/AI-EXPRESS/deploy/ -rf
+ cp /opt/AI-EXPRESS/output/body_solution /opt/AI-EXPRESS/deploy/ -rf
+ cp /opt/AI-EXPRESS/output/face_body_multisource /opt/AI-EXPRESS/deploy/ -rf
+ cp /opt/AI-EXPRESS/output/xwarehouse_sample /opt/AI-EXPRESS/deploy/ -rf
cp: 无法获取'/opt/AI-EXPRESS/output/xwarehouse_sample' 的文件状态(stat): No such file or directory
+ mkdir -p /opt/AI-EXPRESS/deploy/ssd_test/config/vio_config
+ mkdir -p /opt/AI-EXPRESS/deploy/ssd_test/config/bpu_config
+ cp /opt/AI-EXPRESS/output/face_solution/configs/bpu_config.json /opt/AI-EXPRESS/deploy/ssd_test/config/bpu_config
+ mkdir -p /opt/AI-EXPRESS/deploy/ssd_test/config/models
+ cp -r '/opt/AI-EXPRESS/models/x3/ssd/so/*' /opt/AI-EXPRESS/deploy/ssd_test/config/models
cp: 无法获取'/opt/AI-EXPRESS/models/x3/ssd/so/*' 的文件状态(stat): No such file or directory
+ cp -r /opt/AI-EXPRESS/source/solution_zoo/xstream/methods/ssd_method/config/ssd_module.json /opt/AI-EXPRESS/source/solution_zoo/xstream/methods/ssd_method/config/ssd_test_workflow.json /opt/AI-EXPRESS/deploy/ssd_test/config
+ '[' x3 == x3 ']'
+ cp -r /opt/AI-EXPRESS/deploy/configs/hb_camera_x3.json /opt/AI-EXPRESS/deploy/configs/hb_vio_x3_1080_fb.json /opt/AI-EXPRESS/deploy/configs/hb_vio_x3_1080.json /opt/AI-EXPRESS/deploy/configs/hb_vio_x3_608.json /opt/AI-EXPRESS/deploy/ssd_test/config/vio_config
+ cp -r /opt/AI-EXPRESS/deploy/configs/vio /opt/AI-EXPRESS/deploy/ssd_test/config/vio_config
+ cp /opt/AI-EXPRESS/output/video_box /opt/AI-EXPRESS/deploy/ -rf
+ cp /opt/AI-EXPRESS/output/video_box/data/test.264 /opt/AI-EXPRESS/deploy/ -rf
+ cp -r /opt/AI-EXPRESS/build/bin/ssd_method_test /opt/AI-EXPRESS/deploy/ssd_test/
+ cp -r /opt/AI-EXPRESS/source/solution_zoo/xstream/methods/ssd_method/test/data /opt/AI-EXPRESS/deploy/ssd_test
+ mkdir -p /opt/AI-EXPRESS/deploy/python_api
+ mkdir -p /opt/AI-EXPRESS/deploy/python_api/tests
+ mkdir -p /opt/AI-EXPRESS/deploy/configs/pytest_configs
+ cp /opt/AI-EXPRESS/source/common/xstream/python_api/package/xstream /opt/AI-EXPRESS/deploy/python_api/ -rf
+ cp /opt/AI-EXPRESS/source/common/xproto/python_api/package/xproto /opt/AI-EXPRESS/deploy/python_api/ -rf
+ cp /opt/AI-EXPRESS/source/common/xstream/python_api/package/tests/test_session.py /opt/AI-EXPRESS/source/common/xstream/python_api/package/tests/test_xstream1.py /opt/AI-EXPRESS/source/common/xstream/python_api/package/tests/test_xstream2.py /opt/AI-EXPRESS/deploy/python_api/tests/ -rf
+ cp /opt/AI-EXPRESS/source/common/xproto/python_api/package/tests/test_async.py /opt/AI-EXPRESS/deploy/python_api/tests/ -rf
+ cp /opt/AI-EXPRESS/source/common/xstream/python_api/package/configs/a_filter.json /opt/AI-EXPRESS/source/common/xstream/python_api/package/configs/b_filter.json /opt/AI-EXPRESS/deploy/configs/pytest_configs/ -rf
+ cp /opt/AI-EXPRESS/webservice /opt/AI-EXPRESS/deploy/ -rf
+ mkdir -p /opt/AI-EXPRESS/deploy/vioplugin_test
+ cp -r /opt/AI-EXPRESS/build/bin/vioplugin_sample /opt/AI-EXPRESS/deploy/vioplugin_test/
[zjh@localhost AI-EXPRESS]$
无法获取是啥意思啊
2020-12-26
2
13
- yanghangjun回复张建宏:
  请问你的版本是哪里获取的。 github上还未开源手势识别的solution。
  若是地平线提供的版本，可以告知下版本信息，我们复现下
  2020-12-28
  0
  回复
- 张建宏回复yanghangjun:
  开发板的镜像是目前最新的镜像
  x3sdbx3-samsung2G
  2020-12-28
  0
  回复
- 张建宏回复yanghangjun:
  打包的时候报了这个错，之后上板运行也出现了错误，你们可以看下我发的另外一个帖子
  2020-12-28
  1
  回复
- 张建宏回复yanghangjun:
  2020-12-28
  1
  回复
- 张建宏回复yanghangjun:
  AI-EXPRESS v2.7.0
  2020-12-28
  0
  回复
- 张建宏回复yanghangjun:
  这个AI-EXPRESS不都是git下来的吗，我就git下来的AI-EXPRESS 2.7.0 开发机环境centos7.6
  2020-12-28
  0
  回复
- sailor回复张建宏:
  AI-EXPRESS 2.7.0版本目前不支持20201124的镜像，2.7.0版本，请使用20201023的镜像。
  2020-12-28
  1
  回复
- 张建宏回复sailor:
  我在开发机centos7.6上打包出现了错误，这和板子的镜像无关吧
  2020-12-28
  2
  回复
- 张建宏回复sailor:
  也就是说我把v2.70 的deploy包放到20201124的板子上是运行不了的
  2020-12-28
  0
  回复
- sailor回复张建宏:
  在开发机centos7.6上打包出现的错误和板子的镜像无关，是因为部分参考方案未正式开源而报的错误，这些报错不会影响到已开源的参考方案的正常运行，建议忽略掉他们即可。
  2020-12-28
  0
  回复
- sailor回复张建宏:
  是的，V2.7.0的deploy包应该对应的系统镜像是20201023版本。
  ·
  2020-12-28
  0
  回复
- 张建宏回复张建宏:
  安装了之前的镜像已经成功运行，感谢感谢
  2020-12-28
  0
  回复
- sailor回复张建宏:
  好的，不用谢。
  2020-12-28
  0
  回复
乔纳森
Lv.1
有成功运行，可以展示的demo视频效果吗？
2020-08-14
0
3
- yanghangjun回复乔纳森:
  AI-EXPRESS 2.4.0版本包中，有一个手势识别的示例，里面串联了手体检测、手关键点检测以及手势识别，支持实时画面输入，WEB展示端看检测以及手势识别结果
  2020-08-14
  0
  回复
- 张建宏回复yanghangjun:
  手势识别的示例该如何运行下呢
  2020-12-28
  0
  回复
- 张建宏回复yanghangjun:
  2.4的包里有手势识别的案例嘛？
  2021-01-06
  0
  回复
张建宏
Lv.1
AI-Express 2.4.0中有手势识别的参考案例吗？如何下载？
2021-03-04
2
0
张建宏
Lv.1
能提供一下2.4.0版本的开发包吗
2021-06-02
2
0
linjinbin
Lv.1
赞赞赞！?
2020-08-22
0
0
张建宏
Lv.1
板子模型运行了，但打开摄像头失败
2020-12-26
0
0