ML models to support tasks such as NSFW detection for imagery associated with the Wikimedia projects is not a new task / request nor is it a straightforward task (there are many different approaches that one might take). This task will cover the work to gather information around the following aspects and make it easily accessible (likely on meta) for this project or any future investigations:
- Past attempts to build similar models
- Concerns raised in the past about the potential models and how they might be applied
- Approaches taken for this modeling on other platforms and issues that have arisen
- Different potential taxonomies -- e.g., binary yes/no vs. more nuanced types of imagery