ML-Deployment 📦

A simple example of deploying machine learning/deep learning applications.

Tech Stack

In this example,

• A PyTorch-based object detector is built
• Served using Flask Endpoints
• The tasks are queued and called asynchronous with Celery
• The Job history and the results are saved in MongoDB
• Deployed using Docker and Docker-Compose

Requirements

• Docker
• docker-compose
• nvidia-docker2

Installation

Build the docker image:

docker-compose build

Start and attach to the docker-compose services:

docker-compose up

Endpoints

• Start a Job : POST /api/v1/jobs

  curl -X POST -H "Content-Type: application/json" -d '{"url":"http://ai.stanford.edu/~jkrause/cars/car2.jpg"}' http://127.0.0.1:5000/api/v1/jobs
  

  Response:

  {
      "id": "60ad59866dc08443a331a030", 
      "result": [],
      "source": "http://ai.stanford.edu/~jkrause/cars/car2.jpg", 
      "state": "Initialized", 
      "time_created": "2021-05-25 20:09:42.542180"
  }
  

• Get all Jobs : GET /api/v1/jobs

  curl -X GET http://127.0.0.1:5000/api/v1/jobs
  

  Response:

  [
       {
          "id": "60ac55a79d86cb024078a112", 
          "result": [
          {
              "box": {
                  "x1": 0.04, 
                  "x2": 0.54, 
                  "y1": 0.55, 
                  "y2": 0.96
              }, 
              "label": "cat", 
              "score": 0.91
          }, 
          {
              "box": {
                  "x1": 0.62, 
                  "x2": 0.82, 
                  "y1": 0.65, 
                  "y2": 0.99
              }, 
              "label": "person", 
              "score": 0.8
          }, 
          {
              "box": {
                  "x1": 0.09, 
                  "x2": 0.14, 
                  "y1": 0.14, 
                  "y2": 0.28
              }, 
              "label": "person", 
              "score": 0.8
          }, 
          {
              "box": {
                  "x1": 0.8, 
                  "x2": 0.95, 
                  "y1": 0.59, 
                  "y2": 0.99
              }, 
              "label": "person", 
              "score": 0.8
          }, 
          {
              "box": {
                  "x1": 0.14, 
                  "x2": 0.45, 
                  "y1": 0.16, 
                  "y2": 0.4
              }, 
              "label": "car", 
              "score": 0.72
          }, 
          {
              "box": {
                  "x1": 0.43, 
                  "x2": 0.52, 
                  "y1": 0.16, 
                  "y2": 0.25
              }, 
              "label": "truck", 
              "score": 0.65
          }, 
          {
              "box": {
                  "x1": 0.34, 
                  "x2": 0.38, 
                  "y1": 0.14, 
                  "y2": 0.23
              }, 
              "label": "person", 
              "score": 0.65
          }, 
          {
              "box": {
                  "x1": 0.3, 
                  "x2": 0.34, 
                  "y1": 0.15, 
                  "y2": 0.22
              }, 
              "label": "person", 
              "score": 0.62
          }, 
          {
              "box": {
                  "x1": 0.36, 
                  "x2": 0.43, 
                  "y1": 0.16, 
                  "y2": 0.24
              }, 
              "label": "car", 
              "score": 0.57
          }
          ], 
          "source": "https://1.bp.blogspot.com/-yuodfZa6gyM/XlbQfiAzbzI/AAAAAAAAFYA/QSTnuZksQII2PaRON2mqHntZBHL-saniACLcBGAsYHQ/s1600/Figure1.png", 
          "state": "Done", 
          "time_created": "2021-05-25 01:40:55.027000"
      }, 
  
      {
          "id": "60ad59866dc08443a331a030", 
          "result": [
          {
              "box": {
                  "x1": 0.19, 
                  "x2": 0.73, 
                  "y1": 0.48, 
                  "y2": 0.89
              }, 
              "label": "car", 
              "score": 0.88
          }, 
          {
              "box": {
                  "x1": 0.58, 
                  "x2": 0.62, 
                  "y1": 0.52, 
                  "y2": 0.59
              }, 
              "label": "person", 
              "score": 0.73
          }
          ], 
          "source": "http://ai.stanford.edu/~jkrause/cars/car2.jpg",
          "state": "Done", 
          "time_created": "2021-05-25 20:09:42.542000"
      }
  ]
  

• Get a single Job : GET /api/v1/jobs/<job_id>

  curl -X GET http://127.0.0.1:5000/api/v1/jobs/60ad413cc91fb4916a53315f
  

  {
      "id": "60ad59866dc08443a331a030", 
      "result": [
      {
          "box": {
              "x1": 0.19, 
              "x2": 0.73, 
              "y1": 0.48, 
              "y2": 0.89
          }, 
          "label": "car", 
          "score": 0.88
      }, 
      {
          "box": {
              "x1": 0.58, 
              "x2": 0.62, 
              "y1": 0.52, 
              "y2": 0.59
          }, 
          "label": "person", 
          "score": 0.73
      }
      ], 
      "source": "http://ai.stanford.edu/~jkrause/cars/car2.jpg",
      "state": "Done", 
      "time_created": "2021-05-25 20:09:42.542000"
  }