Narrow down by specifying conditions
現在1889件がヒットしています。check
What is Pose Estimation?
Pose estimation is a technology for understanding human movement and posture. By combining computer vision and deep learning, images/It becomes possible to identify each part of the human body in the video and connect them to analyze the overall posture and movement. This makes it possible to understand human movement patterns and to estimate posture in real time. It is expected to be applied to a variety of applications such as safety monitoring, sports, medical rehabilitation, and entertainment.
[This demo]
On this page,IoT/Released for embedded devices “Qualcomm Snapdragon”“Pose estimationA.I.We are conducting a demonstration that incorporates the model (*). You can also watch the video generated by executing the command.
*Google CoralPublished in “PoseNet”model(Based on MobileNetV1 architecture)use.
PoseNet can detect 17 key points. Key points are concepts that play a central role in pose estimation technology. Refers to the individual parts that make up human posture, such as the right eye, left ear, right wrist, left knee, right ankle, etc. Keypoints include both locations (coordinates) and confidence scores (a measure of how accurately the area was detected) and cover major joints and distinctive areas of the human body. This allows an analysis of the overall posture.
Pose estimation using PoseNet
testing environment
Evaluation board withQualcomm QCS6490(C6490 Development Kit) (Jump to product page)・
- OS: Ubuntu 20.04 (SDKversion: C6490.LU1.0.CS.r002002)
・Windows PC (command prompt)
・USB Type-A to Type-Ccable
Environment setup
・Google Coralis published byGithubfromPoseNetDownload the model.
"posenet_mobilenet_v1_075_481_641_quant.tflite"I used
・Prepare a label file. This time,we prepared a file called"posenet.labels".
・We will prepare a video that will be processed when the demo is executed. This time,I prepared a file called"dance.mp4".
Transfer each file to the C6490 Development Kit.
(PC) $ adb push posenet_mobilenet_v1_075_481_641_quant.tflite /data/TFLite
(PC) $ adb push posenet.labels /data/TFLite
(PC) $ adb push dance.mp4 /data/misc/cameraExecution of pose estimation process
Prepared using the above steps using this command.mp4Pose estimation processing is performed using the file video. Draw a line connecting key points detected in the video and create a newmp4Generate the file. In the example below, a video file called "camera_posenet.mp4" will be saved.
(C6490) # gst-launch-1.0 -e qtivcomposer name=mixer sink_1::dimension="<1280,720>" ! queue ! qtic2venc ! h264parse ! mp4mux ! queue ! filesink location="/data/camera_posenet.mp4" filesrc location=/data/misc/camera/dance.mp4 ! qtdemux name=demux. ! queue ! h264parse ! qtivdec ! video/x-raw\(memory:GBM\) ! queue ! tee name=split ! queue ! mixer. split. ! queue ! qtimlvconverter ! queue ! qtimltflite delegate=hexagon model=/data/TFLite/posenet_mobilenet_v1_075_481_641_quant.tflite ! queue ! qtimlvpose threshold=51.0 results=2 module=posenet labels=/data/TFLite/posenet.labels ! video/x-raw,format=BGRA,width=640,height=360 ! queue ! mixer.Let's play the video generated by executing the command.
On this page,we tried running the pose estimation model PoseNeton theTurboXC6490 Development Kitequipped withQualcomm QCS6490 SoC.
QualcommCompany's “Snapdragon series”teeth,IoT/For embedded devicesSoCWe have a large lineup of image processing, such as the examples introduced on this page.A.I.Application use cases are also increasing. This time I used “QCS6490”is a middle class product, but we also have other products that can be proposed for a wide range of uses, from entry class to premium class. If you would like more information, please contact us.
Inquiry/Quotation
If you have any questions about the contents of this page or would like detailed product information, please contact us here.
