Models Deployment

Deploying a application is relatively easy.

There are several ways to do it depending on your specific use case and the tools that you use.

You will see more about some of the ways to do it in the next sections.

Versions

The API is already being used in production in a few applications and systems. But its development is still moving quickly.

New features are added frequently, bugs are fixed regularly, and the code is still continuously improving.

That's why the current versions are still 0.x.x, this reflects that each version could potentially have breaking changes. This follows the Semantic Versioning conventions.

You can create production applications with this API implementation right now (and you have probably been doing it for some time), you just have to make sure that you use a version that works correctly with the rest of your code.

Requirements

Python 3.6+

Quickstart

In the root project execute the following command to install all dependencies project

pip install -r requirements.txt

You will also need an ASGI server, for production such as Uvicorn or Hypercorn.

pip install uvicorn

Configuration

Configuration Files

All of the configuration files for the API are stored in the config directory. Each option is documented, so feel free to look through the files and get familiar with the options available to you.

Application Key

The next thing you should do after installing is set your application key to a random string.

Typically, this string should be 32 characters long. The key can be set in the .env environment file. If you have not copied the .env.example file to a new file named .env, you should do that now. If the application key is not set, your user sessions and other encrypted data will not be secure!

Model Configuration

The model that we are going to deploy is for predicting photovoltaic fault. You can get the data here.

Model	Weights Trained	Config
SSD7 Panel	weight	config
SSD300 Soiling	weight	config
YOLO3 Panel	weight	config
YOLO3 Soiling	weight	config
YOLO3 Diode	weight	config
YOLO3 Affected Cell	weight	config

We start by loading the data and compiled models into the storage/model folder and the configuration files for each model in the storage/config folder and saving the names of the features that we want to use in our model.

Example model configuration file

{
  "model": {
    "min_input_size": 400,
    "max_input_size": 400,
    "anchors": [5,7, 10,14, 15, 15, 26,32, 45,119, 54,18, 94,59, 109,183, 200,21],
    "labels": ["1"],
    "backend": "full_yolo_backend.h5"
  }
}

After we have prepared the data and saved all necessary files it is time to start creating the API to serve our model from.

NOTE: There are several methods for saving a model, each with its own sets of pros and cons you may change in function of your necessity.

Run it

Run the server with:

uvicorn run:app --reload

About the command uvicorn run:app --reload...

The command uvicorn run:app refers to:

• run: the file run.py (the Python "module").
• app: the object imported inside of run.py with the line from app import app.
• --reload: make the server restart after code changes. Only do this for development.

You already created an API that:

• Receives HTTP requests in the path /prediction.
• /prediction path take POST operations (also known as HTTP methods).
• The path parameter should be parse as json object.
• If the mimetype does not indicate JSON application/json this returns None.

Example of an input data

[
  {
    "id": 3,
    "url": "https://domain.com/storage/3/0074.jpg"
  },
  {
    "id": 6,
    "url": "https://domain.com/storage/6/0045.jpg"
  },
  {
    "id": 7,
    "url": "https://domain.com/storage/7/0055.jpg"
  },
  {
    "id": 8,
    "url": "https://domain.com/storage/8/0024.jpg"
  },
  {
    "id": 2,
    "url": "https://domain.com/storage/2/0078.jpg"
  },
  {
    "id": 5,
    "url": "https://domain.com/storage/5/0091.jpg"
  },
  {
    "id": 4,
    "url": "https://domain.com/storage/4/0089.jpg"
  }
]

Viewing Results

[
  {
    "id": 3,
    "url": "https://domain.com/storage/3/0074.jpg",
    "objects": []
  },
  {
    "id": 6,
    "url": "https://domain.com/storage/6/0045.jpg",
    "objects": [
      {
        "class": "1",
        "label": "Soiling Fault",
        "score": -1,
        "xmax": 98,
        "xmin": 86,
        "ymax": 100,
        "ymin": 100
      },
      {
        "class": "1",
        "label": "Soiling Fault",
        "score": -1,
        "xmax": 256,
        "xmin": 243,
        "ymax": 104,
        "ymin": 104
      },
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 14,
        "xmin": 0,
        "ymax": 49,
        "ymin": 49
      },
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 231,
        "xmin": 218,
        "ymax": 56,
        "ymin": 56
      },
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 231,
        "xmin": 218,
        "ymax": 59,
        "ymin": 59
      }
    ]
  },
  {
    "id": 7,
    "url": "https://domain.com/storage/7/0055.jpg",
    "objects": []
  },
  {
    "id": 8,
    "url": "https://domain.com/storage/8/0024.jpg",
    "objects": []
  },
  {
    "id": 2,
    "url": "https://domain.com/storage/2/0078.jpg",
    "objects": [
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 364,
        "xmin": 347,
        "ymax": 329,
        "ymin": 329
      },
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 367,
        "xmin": 343,
        "ymax": 329,
        "ymin": 329
      },
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 366,
        "xmin": 350,
        "ymax": 329,
        "ymin": 329
      },
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 369,
        "xmin": 347,
        "ymax": 329,
        "ymin": 329
      },
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 361,
        "xmin": 347,
        "ymax": 334,
        "ymin": 334
      },
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 366,
        "xmin": 349,
        "ymax": 333,
        "ymin": 333
      },
      {
        "class": "4",
        "score": -1,
        "label": "Diode Fault",
        "xmax": 367,
        "xmin": 349,
        "ymax": 331,
        "ymin": 331
      }
    ]
  },
  {
    "id": 5,
    "url": "https://domain.com/storage/5/0091.jpg",
    "objects": []
  },
  {
    "id": 4,
    "url": "https://domain.com/storage/4/0089.jpg",
    "objects": []
  }
]

FAQs

Current version

Important: The default version of the API may change in the future. If you're building an application and care about the stability of the API, be sure to fork the master branch.

Media types

The Media Type is specified in header of request. The most basic media types the API supports are:

application/json

Neither of these specify a version, so you will always get the current default JSON representation of resources.

For any media type format that is not supported, the Api should return a 406 Not Acceptable status code.

Troubleshooting

If you're encountering some oddities in the API, here's a list of resolutions to some of the problems you may be experiencing.

• Why am I getting a 404?

The request could not be understood by the server due to malformed syntax. The client should not repeat the request without modifications

• Why am I not seeing all my results?

Most API calls accessing a list of resources (e.g., users, issues, etc.). If you're making requests and receiving an incomplete set of results, a response is specified in an unsupported content type.

• Why am I getting a 500?

Server Mistake - Indicates that something went wrong on the server that prevent the server from fulfilling the request.

Third-party libraries

• Flask
• TensorFlow
• NumPy
• pandas