Example-YOLO

This is an example of service written in Python, it provides a base from which make other services related to object detection using YOLOv8 from Ultralytics. The implementation uses gunicorn, Flask

Installation

To run this service locally, you require Python which can be downloaded here and you may require the Docker daemon as well, which can be easily installed with Docker Desktop. However, if you plan to run this service on the cloud, the best way is to fork this repo and edit the cd.yml file to deploy it with Continous Deployment. (CD) Also, after forking you can set the CD with your provider, it must be possible for sure.

Following the local installation, run the following commands in order:

git clone https://github.com/Vicente-G/Example-YOLO
cd example-yolo

Now you should be on the service folder, from where you can run it with either Python or Docker, depending on your needs.

Running

If you plan on editing this service to your needs, you may prefer Flask, as it can be continously tested with its debug mode. To run this example using Python, install the dependencies and run the dev script, like this:

pip install --user pdm
pdm sync
pdm run dev

In the other hand, if you plan to just run the service in your local server, the best option is to use containers, for reliability and security. With that on mind, Docker is your option. First enable the Docker daemon, and right after run these commands:

docker build -t example-yolo .
docker run -p 8080:8080 -e PORT='8080' example-yolo

In any way you run the service, it will block your terminal. Now, to test its responses, you can try any HTTP request emitter, such as Postman, Axios or cURL. The following example shows a GET request over the service with cURL

curl --head localhost:8080/status

If everything is working as intended, the answer to that response should be a 200, which means that everything is okay! (The 200 should be in the first line or so)

Requests

As this is just an example, this service only has one extra route asides from the status one. This service provides the route process, with the following command:

curl -X POST localhost:8080/process \
-d '{"image":"<your Base64 string encoding an image>"' \
-H 'Content-Type: application/json'

The output should be something with the following shape: (key changes to error if the processing fails)

{
    "image": "<a Base64 string encoding the labelled image>",
    "locs": "[A number array with the mid points of the boxes]",
    "classes": "[A number array with the classes detected]"
}

Also, the following correspond to the equivalent action in axios:

const { error, ...output } = await axios.post(
    'localhost:8080/process',
    { image: '<your Base64 string encoding an image>' },
    {
        headers: {
            'Content-Type': 'application/json'
        }
    }
)