7.8 billion different people but just five conventional t-shirt sizes – XS, S, M, L, and XL. Hence, to provide a much better fit, we are creating three different t-shirt lengths (the Short, Regular, and Long) for each conventional t-shirt size – providing you with fifteen different options. The following table clearly explains the sizes provided.

To further ease this process, the following Neural Network powered API takes the user’s height, weight, shoe size, and preferred t-shirt fit (Standard or Relaxed) and provides the perfect t-shirt with its respective confidence percentage.

Please note that this API is currently in its development stage and can only consider men’s sizes. Future developments include reformatting the front end of the API and improving the model to provide more accurate t-shirt sizes.

The following visual illustrates the working of the API.

To run the program, clone the directory using the following code.

git clone https://github.com/shantam-8/Machine-Learning-API-for-Tshirt-Size-Determination.git

Then, install the dependencies by running the followinng code.

 pip install -r requirements.txt

You are set to go. Run the following code to execute the program.

python app.py

Click on http://127.0.0.1:8000/predict after the following is seen on the terminal.

INFO:     Started server process [22572]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)

The model (Model_4.h5) is a Neural Network constructed using the TensorFlow Library. After several evaluations it was adjudged that the “swish” activation increased the accuracy of the model when compared with standard activations like “ReLU”. The "swish" activation was used in the following manner.

#Function to delcare swish as an activation.
from keras.backend import sigmoid
from keras.utils.generic_utils import get_custom_objects
from keras.layers import Activation

def swish(x, beta = 1):
    return (x * sigmoid(beta * x))

get_custom_objects().update({'swish': swish})

#Replacing "ReLU" with "swish" in the activation parameter of Sequential. 
import tensorflow as tf

model = tf.keras.models.Sequential([
                                    tf.keras.layers.Dense(6000, activation="swish"),
                                    ])

Additionally, of the 500 data points sourced from different websites, 70% were used as the training set and 30% were used as the test set. Through several processes of trial-and-error, a Training Accuracy of 0.9521 and a Test Accuracy of 0.8541 was achieved.

As the original model was too large, the model has been compressed into a model named "Model_4.tflite" and suitable changes have been made to "app.py". This is seen in the "deploy" branch of the repository. This branch has been deployed via heroku. The following site opens the API: https://tshirt-sizing.herokuapp.com.

The list indicates the updates that can be made to this API.

• Fixing the Heroku deployment.
• Increasing the accuracy of the model.
• Improving the front-end of the API.
• Updating the front-end and back-end of the API to allow values for female t-shirts.

• "ML - Swish Function by Google in Keras.” GeeksforGeeks, 26 May 2020, https://www.geeksforgeeks.org/ml-swish-function-by-google-in-keras/.
• “How to Deploy Machine Learning/Deep Learning Models to the Web.” KDnuggets, https://www.kdnuggets.com/2021/04/deploy-machine-learning-models-to-web.html.
• “Delply Machine Learning Model Using Heroku and FASTAPI.” Analytics Vidhya, 6 July 2021, https://www.analyticsvidhya.com/blog/2021/06/deploying-ml-models-as-api-using-fastapi-and-heroku/.

This repository has a MIT License.