Edge ML Filter Plugin

The plugin takes a image saved by south plugin named webcam media, sends that to Edge ML cluster running somewhere else. The Edge ML cluster returns a response in the form of json which contains information about detected objects, their bounding boxes and confidence score. This information is appended to the readings generated from south service.

• ‘assetName’: type: string default: ‘edgeMl’:

  Name of asset to listen on; readings have path names of images to analyze. The plugin will pick these path names to read these images.

• ‘outAssetName’: type: string default: ‘edgeMlInference’:

  Name of asset to write ML inferences on.

• ‘deploymentName’: type: string default: ‘’:

  Name of Kubernetes deployment for ML model

• ‘edgeMlUrl’: type: string default: ‘’:

  REST URL for ML model which analyzes images; dynamically discovered if empty.

• ‘forwardData’: type: boolean default: ‘true’:

  Forward data as well as inferences.

• ‘rmFile’: type: string default: ‘false’:

  Remove source files after inference.

• ‘enable’: type: boolean default: ‘true’:

  Enable/ Disable the plugin.

Installation

To run the plugin you must follow these prerequisites.

1. Run the south webcam media plugin.

   To run the south webcam media plugin you can either

   1. Copy some images inside some directory in FOGLAMP_ROOT/data. Let’s say the directory name is pics. Run the following command.

      curl -sX POST http://localhost:8081/foglamp/service -d '{"name":"My_web_cam","type":"south","plugin":"webcam_media","enabled":false,"config":{"assetName":{"value":"WebcamImages"}, "imageDir":{"value":"pics"}, "mediaType":{"value":"directory"}, "fpm":{"value":"10.0"}}}' |
      

   2. Connect a camera to the machine and run the following command.

      $ v4l2-ctl --list-formats-ext --device /dev/video0
      You will see something like
      'YUYV' (YUYV 4:2:2)
        Size: Discrete 640x480
                Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 720x480
                Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 1280x720
                Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 1920x1080
                Interval: Discrete 0.067s (15.000 fps)
                Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 2592x1944
                Interval: Discrete 0.067s (15.000 fps)
        Size: Discrete 0x0
      

      Now we know that the id 0 is functional. If no output then try 1,2,3 and so on.

      Finally launch the plugin using

      curl -sX POST http://localhost:8081/foglamp/service -d '{"name":"My_web_cam","type":"south","plugin":"webcam_media","enabled":false,"config":{"assetName":{"value":"WebcamImages"}, "imageDir":{"value":"webcam"}, "mediaType":{"value":"camera"}, "cameraNumber":{"value":"0"}, "fpm":{"value":"10.0"}}}' |jq
      

2. Start the Edge ML cluster.

   For starting the Edge ML cluster you should follow this README file.

3. Add the filter Edge ML.

curl -sX POST http://localhost:8081/foglamp/filter -d '{"name":"edge_ml_filter","plugin":"edgeml","filter_config":{"deploymentName":"edgeml-deployment","assetName":"WebcamImages","outAssetName":"DetectionResults","enable":"true", "forwardData":"true", "rmFile":"false"}}' |jq
curl -sX PUT 'http://localhost:8081/foglamp/filter/My_web_cam/pipeline?allow_duplicates=true&append_filter=true' -d '{"pipeline":["edge_ml_filter"]}' |jq

4. Finally Enable the schedule.

curl -sX PUT http://localhost:8081/foglamp/schedule/enable -d '{"schedule_name":"My_web_cam"}' |jq

See Also

foglamp-filter-ADM_LD_prediction - Filter to detect whether a large discharge is required for a centrifuge

foglamp-filter-breakover - Filter to forecast the possibility of a breakover.

foglamp-filter-greyscale - Convert 24bit RGB images to greyscale images

foglamp-filter-image-bounding-box - Uses the output of an ML model to crop out or draw bounding boxes on an image.

foglamp-filter-image-resize - Python based plugin to resize images by scaling or cropping.

foglamp-filter-mirror - A filter plugin to mirror image type data points

foglamp-filter-rotate - Rotate all images found in datapoints within a reading

foglamp-filter-vibration_features - A filter plugin that takes a stream of vibration data and generates a set of features that characterise that data

foglamp-north-png - A plugin to write an image type data points to PNG files in the local filesystem

foglamp-south-video4linux - A south plugin to ingests images from various devices using the Video4Linux API. Video4Linux supports a wide variety of video capture devices on Linux platforms.

foglamp-south-webcam_media - A FogLAMP south plugin that forwards image data, either directly from a webcam or from a directory of images