GRPC Yolov5s Pipeline Service

Overview

This EdgeX-based application service provides an interface to use gRPC to call OVMS to run the input data through a specified model. The service may be set up to use different topics to trigger different pipelines and call different models. The current implementation makes a call to OVMS running the Yolov5s model.

Getting Started

Clone the Aicsd code as a submodule in the directory retail-use-cases/use-cases
```
git submodule add https://github.com/intel/AiCSD
```
Change directories to retail-use-cases
```
cd ..
```

Download models and sample media

make download-models download-sample-media

Build the OVMS Docker container
```
make build-ovms-server
```
change directories to use-cases/aicsd
```
cd use-cases/aicsd
```
Build the Pipeline Validator and Pipeline grpc-go services
```
make docker-pipeline-val docker-pipeline-grpc-go
```

Modify docker-compose-pipeline-val.yml evnironment variables to have the following settings

APPLICATIONSETTINGS_PIPELINEHOST: pipeline-grpc-go
APPLICATIONSETTINGS_PIPELINEPORT: 59790

Run the pipeline validator and pipeline grpc-go services along with their dependencies
```
make run-pipeline-grpc-go
```

Verify that the pipeline-grpc-go container has started by checking the log for the message

level=INFO ts=2024-05-30T20:06:58.037173489Z app=app-pipeline-grpc-go source=messaging.go:125 msg="Waiting for messages from the MessageBus on the 'ovms-grpc/yolov5' topic"

Send a POST request to http://localhost:59788/api/v1/launchPipeline with the following body

{
    "InputFileLocation":"rtsp://camera-simulator:8554/camera_0",
    "PipelineTopic":"ovms-grpc/yolov5",
    "OutputFileFolder":"0.0.0.0:8555"
}

Verify that the pipeline-grpc-go container has output in its log. This shows that the pipeline is running.

level=INFO ts=2024-05-30T20:06:58.037173489Z app=app-pipeline-grpc-go source=messaging.go:125 msg="Waiting for messages from the MessageBus on the 'ovms-grpc/yolov5' topic"
level=INFO ts=2024-05-30T20:08:37.753137304Z app=app-pipeline-grpc-go source=functions.go:265 msg="RunOvmsModel: Processing time: 45 ms; fps: 16.70864819479429\n"
level=INFO ts=2024-05-30T20:08:37.685039109Z app=app-pipeline-grpc-go source=functions.go:265 msg="RunOvmsModel: Processing time: 37 ms; fps: 16.72014862354332\n"
level=INFO ts=2024-05-30T20:08:40.689475521Z app=app-pipeline-grpc-go source=functions.go:265 msg="RunOvmsModel: Processing time: 49 ms; fps: 16.368045264717768\n"
level=INFO ts=2024-05-30T20:08:40.801842876Z app=app-pipeline-grpc-go source=functions.go:265 msg="RunOvmsModel: Processing time: 29 ms; fps: 16.37919507955609\n"

To stop the data stream from coming in, stop the camera-simulator and camera-simulator0 containers.

Verify that the pipeline-grpc-go container has successfully finished processing by verifying the log says

level=DEBUG ts=2024-05-30T20:10:56.351778319Z app=app-pipeline-grpc-go source=triggermessageprocessor.go:196 msg="trigger successfully processed message 'OVMS Pipeline' in pipeline 64a47e72-3e89-452e-9010-2d64c059c108"

Verify that the job status from the pipeline validator service making a GET request to http://localhost:59788/api/v1/job which should return an entry as follows

[
    {
        "Id": "0",
        "Owner": "none",
        "InputFile": {
            "Hostname": "gateway",
            "DirName": "rtsp://camera-simulator:8554/",
            "Name": "camera_0",
            "Extension": "",
            "ArchiveName": "",
            "Viewable": "",
            "Attributes": {}
        },
        "PipelineDetails": {
            "TaskId": "de9ee7bd-aacb-4e6a-ac91-a3c3be3b09c3",
            "Status": "PipelineComplete",
            "QCFlags": "passed",
            "OutputFileHost": "",
            "OutputFiles": null,
            "Results": "ovms-server0:9001"
        },
        "LastUpdated": 1717025357138890553,
        "Status": "Complete",
        "ErrorDetails": null,
        "Verification": 0
    }
]

Repeat by restarting the camera-simulator services and sending the POST request to launchPipeline.

Tearing Down

To tear down the services and clean up any data created, run

make down clean-files clean-volumes