Skip to content

Train-to-Deploy with Azure customvision on Intel powered IoT Edge Device

License

Notifications You must be signed in to change notification settings

psakamoori/intel-rrk-custom-vision

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Azure ML Training and Deployement on Intel Edge device (RRK)

##Pre-requisites:

  • Intel powered Edge device (UP^2) with Ubuntu 16.04
  • USB webcam (/dev/video0)
  • OpenVINO installation with ORT (ONX Runtime) execution provider
  • Python OpenCV

##Folders: #models (models.onnx, labesl.txt)

  • ONNX pre-trained models for Image_classificaiton
  • ONNX pre-trained models for object_detection #src
  • Python Application code for Image Classification and object detection
  • objdet/model.config - Model configuration file

##Testing

Image-classifcation

  • Folder "models/image_classification" has few pre-trained onnx models
  • Execute command: src/imgcls$ python3 onnx_image_classifciation.py model.config
  • Expected Output: Predicted image classification result with label
  • ##Example:model.config (for Image Classification)
    {
    "Network":0,
    "modeltype":"onnx model",
    "Input":"cam",
    "display":1,
    "mean_vec":[0.485, 0.456, 0.406],
    "stddev_vec":[0.229, 0.224, 0.225],
    "ScaleWidth":224,
    "ScaleHeight":224,
    "InputFormat":"RGB",
    "Runtime":1,
    "MODEL_FILENAME":"../../models/image_classification/dog_n_cat/model.onnx",
    "LABELS_FILENAME":"../../models/image_classification/dog_n_cat/labels.txt"
    }
    Classification results

Object Detection

  • Folder "models/object_detect/" has face detection pre-trained onnx model
  • Execute command: src/objdet$ python3 onnxruntime_predict.py model.config
  • Expected Output: Renders webcam video frames with inference results (bounding box, detection label and score)
  • ##Example:model.config (for Object detection)
    {
    "Network":0,
    "modeltype":"onnx model",
    "Input":"cam",
    "display":1,
    "Anchors": [[0.573, 0.677], [1.87, 2.06], [3.34, 5.47], [7.88, 3.53], [9.77, 9.17]],
    "ScaleWidth":416,
    "ScaleHeight":416,
    "InputFormat":"RGB",
    "ConfThreshold":0.5,
    "IOU_THRESHOLD":0.45,
    "Runtime":1,
    "MODEL_FILENAME":"../../models/object_detection/face_detect/model.onnx",
    "LABELS_FILENAME":"../../models/object_detection/face_detect/labels.txt"
    }

Face Detection output

Car and Traffic Light Detection

About

Train-to-Deploy with Azure customvision on Intel powered IoT Edge Device

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages