Instructions for future SD teams

Project Overview

Eating disorders (ED) are serious medical conditions exacerbated by increased societal pressures to fit “the thin ideal.” Exposure to images and media that embody this ideal can be triggering to people with ED. Social media platforms are especially rife with these triggers: individuals with ED have created pro-ED communities where they support one another in acting on the symptoms of their ED. These websites consist of inspirational images Therefore, triggering content online poses a serious risk to social media users with ED and those at risk for developing them. Similarly, it is essential that clinicians and family members be able to identify websites containing images that are associated with the promotion of ED to prevent accidental or intentional exposure to these triggers. However, it is challenging for caretakers to find and stay up-to-date with pro-ED communities and content online.

This projects uses deep learning to detect triggering material, with the ultimate goal of designing tools to inform clinicians and support patients in their recovery. The main products of this work are a convolutional neural network that identifies images of ED and two novel software tools built from it that assess websites for ED content. These tools would enable clinicians, family members, and those suffering from ED to understand and identify sources of ED content to improve treatment for ED patients.

Getting started

• see README in the project github repo

Tools and Libraries needed

• Python 2.7
• PyTorch and its dependencies
• a GPU

Pitfalls and issues

What works

• single image URL classification (returns 0 or 1)
• saving and loading models of the classifier across multiple GPUs

Specific moments of breakage/weakness

• mutiple URL classification
• the classifier does not support CPU training or inference

Ideas for next steps

• put in a single URL that has multiple images and classify all of them
• migrating code from (soon-to-be-deprecated) Python 2.7 to Python 3+

---