Skip to content

Latest commit

 

History

History
151 lines (129 loc) · 3.67 KB

README.md

File metadata and controls

151 lines (129 loc) · 3.67 KB
Raw

headout-b

This is the backend server for our Machine Learning as a Service (MLaaS) platform.

Table of Contents

Getting Started

Prerequisites

  • Python
  • Framework Used: Flask
  • pip

Installation

1. Clone this repository:

git clone https://github.com/akshayw1/headout-b.git
cd headout-b

2. Install the dependencies:

pip install -r requirements.txt

Usage

1. Start the Application:

python app.py

The server will start running at http://localhost:5000 by default. PORT number can be changed by specifying the port number at the end of app.py file.

2. API Endpoints

Image Classification

  • POST /upload: Uploads dataset using FormData containing class name and with an image of it, multiple POST request need to be sent to train the model. A simple text response will be sent with the message and status code.

    {
  "folderName": CLASS_NAME,
  "image": IMAGE_URI_STRING
}

    Response:

    {
  "status": 200,
  "message": "Image saved successfully!"
}

  • GET /train: Endpoint to train the model based on given dataset. No body is expected here & a simple status code of 200 is expected on successful training of the model. If no upload data is sent a text message of "No subfolder" with status code 404 will be sent. Response:

    {
    "status": 200
}

  • POST /frame: This will provide the result of live preview of the trained model, when sent with an image frame in FormData. A json response with a status code 200 is expected.

    {
  "image": IMAGE_URI_STRING
}

    Response:

    {
  "status": 200,
  "data": {
    "class1": PROBABILITY_SCORE_OF_CLASS1,
    "class2": PROBABILITY_SCORE_OF_CLASS2,
    "class3": PROBABILITY_SCORE_OF_CLASS3
    .
    .
    .
  }
}

Text Classfication

  • POST /text-train: Uploads dataset using FormData containing a json object for different classes and their examples. Now, model will be trained based on given json data and status code of 200 is expected on successful training of the model

    {
  "data": {
    "class1": [
      SAMPLE_STRING_OF_CLASS1,
      SAMPLE_STRING_OF_CLASS1,
      SAMPLE_STRING_OF_CLASS1
    ],
    "class2": [
      SAMPLE_STRING_OF_CLASS2,
      SAMPLE_STRING_OF_CLASS2,
      SAMPLE_STRING_OF_CLASS2
    ],
    "class3": [
      SAMPLE_STRING_OF_CLASS3,
      SAMPLE_STRING_OF_CLASS3,
      SAMPLE_STRING_OF_CLASS3
    ]
    .
    .
    .
  }
}

    Response:

    {
  "status": 200,
  "message": "Model Trained"
}

  • POST /text-predict: This will provide the result of the trained model, when sent with a text key containing a string to be predicted in FormData. A json response with a status code 200 is expected containing probabilities based on classes of model trained.

    {
  "text": PREDICT_STRING
}

    Response:

    {
  "status": 200,
  "data": {
    "text": STRING_EVALUATED,
    "label_probs": {
      "class1": PROBABILITY_SCORE_OF_CLASS1,
      "class2": PROBABILITY_SCORE_OF_CLASS2,
      "class3": PROBABILITY_SCORE_OF_CLASS3
    .
    .
    .
    }
  }
}