Deploying machine learning models for forensic anthropology with Docker and Streamlit

We're an archeologist and a physicist who created a web application for skeletal sex prediction via machine learning. We knew nothing about web applications, but Docker and Streamlit made our project come to life and now it has important implications for forensic anthropology and bioarchaeology.

In our project, we used two well-known osteoarchaeology datasets to create skeletal sex prediction models based on cranial and long bone measurements. For the analysis of the data and the construction of the models, we used the popular Python libraries pandas and scikit-learn. However, we wanted to go a step further and create a web application, SexEst, to deploy our models online so other scholars can use them for sex prediction in diverse archaeological (and maybe even forensic) assemblages.

After searching for ways to build a web application, we concluded that Streamlit, another Python library, was one of the most straightforward tools and hence we designed and built our web application using Streamlit. A virtual machine running on a remote server was available to us, hence we decided to package all our code and its dependencies inside a Docker image, which we later run as a container in our virtual machine, providing an easy and fast way to have a web application up and running.

In our talk, we'll walk you through the steps we followed in building our web application inside a Docker container. We will present the Dockerfile used, the commands we run to first install Docker to the web server, and then run our image as a Docker container to build and deploy SexEst.

• Speaker :
- Chrysovalantis Constantinou, Computational Scientist
- Efthymia Nikita, Assistant Professor in Bioarchaeology

• References:

--
Join the conversation!