Skip to content

Latest commit

 

History

History
130 lines (109 loc) · 5.87 KB

yaml_configuration_guide.md

File metadata and controls

130 lines (109 loc) · 5.87 KB

Introduction

The contents in .yaml config file should be well structured and follow the supported rules and entity names.

Sample

pipeline_people.yaml

Pipelines:
- name: people
  inputs: [StandardCamera]
  infers: 
    - name: FaceDetection
      model: /opt/openvino_toolkit/models/intel/face-detection-adas-0001/FP16/face-detection-adas-0001.xml
      engine: CPU
      label: /opt/openvino_toolkit/models/intel/face-detection-adas-0001/FP16/face-detection-adas-0001.labels
      batch: 1
      confidence_threshold: 0.5
      enable_roi_constraint: true # set enable_roi_constraint to false if you don't want to make the inferred ROI (region of interest) constrained into the camera frame
    - name: AgeGenderRecognition
      model: /opt/openvino_toolkit/models/intel/age-gender-recognition-retail-0013/FP32/age-gender-recognition-retail-0013.xml
      engine: CPU
      label: to/be/set/xxx.labels
      batch: 16
    - name: EmotionRecognition
      model: /opt/openvino_toolkit/models/intel/emotions-recognition-retail-0003/FP32/emotions-recognition-retail-0003.xml
      engine: CPU
      label: /opt/openvino_toolkit/models/intel/emotions-recognition-retail-0003/FP32/emotions-recognition-retail-0003.labels
      batch: 16
    - name: HeadPoseEstimation
      model: /opt/openvino_toolkit/models/intel/head-pose-estimation-adas-0001/FP32/head-pose-estimation-adas-0001.xml
      engine: CPU
      label: to/be/set/xxx.labels
      batch: 16
  outputs: [ImageWindow, RosTopic, RViz]
  connects:
    - left: StandardCamera
      right: [FaceDetection]
    - left: FaceDetection
      right: [AgeGenderRecognition, EmotionRecognition, HeadPoseEstimation, ImageWindow, RosTopic, RViz]
    - left: AgeGenderRecognition
      right: [ImageWindow, RosTopic, RViz]
    - left: EmotionRecognition
      right: [ImageWindow, RosTopic, RViz]
    - left: HeadPoseEstimation
      right: [ImageWindow, RosTopic, RViz]

Common:

Interface Description

Specify pipeline name

The name value of this pipeline can be anyone other than null.

Specify inputs

Note: The input parameter can only have one value.
Currently, options for inputs are:

Input Option Description Configuration
StandardCamera Any RGB camera with USB port supporting. Currently only the first USB camera if many are connected. inputs: [StandardCamera]
RealSenseCamera Intel RealSense RGB-D Camera, directly calling RealSense Camera via librealsense plugin of openCV. inputs: [RealSenseCamera]
RealSenseCameraTopic Any ROS topic which is structured in image message. inputs: [RealSenseCameraTopic]
Image Any image file which can be parsed by openCV, such as .png, .jpeg. inputs: [Image]
Video Any video file which can be parsed by openCV. inputs: [Video]
IpCamera Any RTSP server which can push video stream. inputs: [IpCamera]

Note: Please refer to this opensource repo RTSP_server_install_guide to install RTSP server for IpCamera input.

Specify input_path

The input_path need to be specified when input is Image, Video and Ipcamera.

Input Option Configuration
Image input_path: to/be/set/image_path
Video input_path: to/be/set/video_path
IpCamera input_path: "rtsp://localhost/test"

Specify infers

The Inference Engine is a set of C++ classes to provides an API to read the Intermediate Representation, set the input and output formats, and execute the model on devices.

  • name

The name of inference engine need to be specified here. Currently, the inference feature list is supported:

Inference Description
FaceDetection Object Detection task applied to face recognition using a sequence of neural networks.
EmotionRecognition Emotion recognition based on detected face image.
AgeGenderRecognition Age and gener recognition based on detected face image.
HeadPoseEstimation Head pose estimation based on detected face image.
ObjectDetection object detection based on SSD-based trained models.
VehicleDetection Vehicle and passenger detection based on Intel models.
ObjectSegmentation object detection and segmentation.
ObjectSegmentationMaskrcnn object segmentation based on Maskrcnn model.

  • model

The path of model need to be specified here. The scheme below illustrates the typical workflow for deploying a trained deep learning model. trained deep learning model

  • engine

Note: Currently, only CPU and GPU are supported.
Target device options are:

Target Device
CPU
Intel® Integrated Graphics
FPGA
Intel® Movidius™ Neural Compute Stick

  • label

Currently, this parameter does not work.

  • batch

Enable dynamic batch size for the inference engine net.

Specify outputs

Note: The output parameter can be one or more.
Currently, the output options are:

Option Description Configuration
ImageWindow Window showing results outputs: [ImageWindow, RosTopic, RViz]
RosTopic Output the topic outputs: [ImageWindow, RosTopic, RViz]
RViz Display the result in rviz outputs: [ImageWindow, RosTopic, RViz]

Specify confidence_threshold

Set the threshold of detection probability.

Specify connects

The topology of a pipe can only have one value on the left and multiple values on the right. The value of the first left node should be the same as the specified inputs.