Deployment

shams3049

I was wondering if it's possible to develop an app that can take a trained model and serve it using docker images, that can be easily configured to run on a remote setup, at which point it can be used to test the model in the real world.

I am starting a research project, that extends to the next two years. as I see supervise as a great asset, and most of the things I need to develop are already quite well done here. I was wondering if anyone can direct me, I can develop the deployment app with this community, and see my time utilized.

max

Hi, this is a nice question. Our current Neural Networks support the deployment as docker containers on any server where Supervisely Agent is installed.

Here are the links to our tutorials and guides:

Supervisely App Engine is relatively new and now we are still in the testing phase. But it provides huge customization/integration opportunities including simplified NN integration (the current NN integration scheme is a bit complicated for many users).

In near future, we will integrate YOLOv5 to the platform using App Engine. It will support training/inference/serving and also can be integrated directly to labeling Interfaces. All sources/tutorials will be available so other users can use them as a start point for their own apps.

shams3049

max Thank you for the pointer, I am super excited about the supervisly vision. Is there a place where you share tasks for the community contribution?

shams3049

max I was wondering, if it's possible to have some direction on how to use the yolov5 serve app using js.

I have formated my request as following, but am getting no where with it, as it shuts down the serve yolov5 app.

var XMLHttpRequest = require("xmlhttprequest").XMLHttpRequest;
var url = "https://app.supervise.ly/public/api/v3/tasks.request.direct";


var xhr = new XMLHttpRequest();
xhr.open("POST", url);
xhr.setRequestHeader("x-api-key", "key goes here");
xhr.setRequestHeader('Content-Type', 'application/json');

image_url = "https://i.imgur.com/tEkCb69.jpg"

xhr.onreadystatechange = function () {
   if (xhr.readyState === 4) {
      console.log(xhr.status);
      console.log(xhr.responseText);
   }};

const json = {
  "taskId": 325348,
  "command": "inference_image_url",
  "data": {
    "image_id": image_url,
  },
};

xhr.send(JSON.stringify(json));
```

my error: 

```
Unexpected exception in main. main_name=main, stack=[
  Traceback (most recent call last):
    File "/usr/local/lib/python3.8/dist-packages/supervisely_lib/function_wrapper.py", line 15, in main_wrapper
      main_func(*args, **kwargs)
    File "/sessions/325348/repo/supervisely/serve/src/sly_serve.py", line 204, in main
      my_app.run(initial_events=[{"command": "preprocess"}])
    File "/usr/local/lib/python3.8/dist-packages/supervisely_lib/app/app_service.py", line 286, in run
      raise self._error
    File "/usr/local/lib/python3.8/dist-packages/supervisely_lib/app/app_service.py", line 191, in handle_message_sync
      self.callbacks[command](api=user_public_api,
    File "/usr/local/lib/python3.8/dist-packages/supervisely_lib/decorators/profile.py", line 11, in wrapper_timer
      value = func(*args, **kwargs)
    File "/sessions/325348/repo/supervisely/serve/src/sly_serve.py", line 113, in inference_image_url
      image_url = state["image_url"]
  TypeError: 'NoneType' object is not subscriptable
```

max

shams3049 Thanks for the kind words. I think we will create a special section here in Ideas Exchange to share our progress and define dev milestones including community contribution.