Defining and launching preconfigured pipelines

Pipelines defined in configuration file can be pre configured i.e. they include all the configurations prior to runtime including any source, destination, parameters etc

These pipelines can be auto started during EVAM start up by setting auto_start flag to true. Refer to this tutorial and this section.

Here is an example of a pre configured pipeline. The pipeline runs a GeTi trained pallet defect detection model on a sample warehouse video. The detection results are written in /tmp/results.jsonl file in JSON format.

{
    "config": {
        "logging": {
            "C_LOG_LEVEL": "INFO",
            "PY_LOG_LEVEL": "INFO"
        },
        "pipelines": [
            {
                "name": "pallet_defect_detection",
                "source": "gstreamer",
                "queue_maxsize": 50,
                "pipeline": "multifilesrc loop=TRUE location=/home/pipeline-server/resources/videos/warehouse.avi ! h264parse ! decodebin ! videoconvert ! gvadetect name=detection model=/home/pipeline-server/resources/models/geti/pallet_defect_detection/deployment/Detection/model/model.xml model-instance-id=inst0 ! queue ! gvawatermark ! gvafpscounter ! gvametaconvert add-empty-results=true name=metaconvert ! gvametapublish name=destination method=file file-path=/tmp/results.jsonl ! appsink name=appsink",
                "auto_start": true
            }
        ]
    }
}

Alternatively, pre-configured pipelines can also be started via REST request (empty) after setting auto_start flag to false

curl localhost:8080/pipelines/user_defined_pipelines/pallet_defect_detection -X POST -H 'Content-Type: application/json' -d '{}'

Providing payload for autostart

In case you have provided placeholders in you pipeline configuration like below, we can also plug in the REST payload in the pipeline configuration and make use of the auto_start feature to start EVAM with this pipeline with the payload provided.

    "pipeline": "{auto_source} name=source  ! decodebin ! videoconvert ! gvadetect name=detection ! queue ! gvawatermark ! gvafpscounter ! gvametaconvert add-empty-results=true name=metaconvert ! gvametapublish name=destination ! appsink name=appsink",
    "parameters": {
        "type": "object",
        "properties": {
            "detection-properties": {
                    "element": {
                    "name": "detection",
                    "format": "element-properties"
                    }
            }
        }
    }

The REST payload can be added inside a payload key section. The autostart is then set true.

The config from above sample will now look like this. Notice auto_start has been set to true and REST payload is provided in the payload section.

    {
        "name": "pallet_defect_detection",
        "source": "gstreamer",
        "queue_maxsize": 50,
        "pipeline": "{auto_source} name=source  ! decodebin ! videoconvert ! gvadetect name=detection ! queue ! gvawatermark ! gvafpscounter ! gvametaconvert add-empty-results=true name=metaconvert ! gvametapublish name=destination ! appsink name=appsink",
        "parameters": {
            "type": "object",
            "properties": {
                "detection-properties": {
                        "element": {
                        "name": "detection",
                        "format": "element-properties"
                        }
                }
            }
        },
        "auto_start": true,
        "payload": {
            "source": {
                "uri": "file:///home/pipeline-server/resources/videos/warehouse.avi",
                "type": "uri"
            },
            "destination": {
                "metadata": {
                    "type": "file",
                    "path": "/tmp/results.jsonl",
                    "format": "json-lines"
                },
                "frame": {
                    "type": "rtsp",
                    "path": "pallet-defect-detection"
                }
            },
            "parameters": {
                "detection-properties": {
                    "model": "/home/pipeline-server/resources/models/geti/pallet_defect_detection/deployment/Detection/model/model.xml",
                    "device": "CPU"
                }
            }
        }
    }

The pipeline would start automatically as soon as EVAM starts.

And of course, we can send new curl requests to launch new instances with different payload parameters.