Hey miss_classified I was pretty disturbed by most of these images. That images was no exception. Honestly till you pointed it out I had no clue. Anyway if it's safe for work or not that's debatable. Would I want my co-workers seeing this on my screen probably not. As I mentioned in the blog right at the end I'm more convinced after going through this exercise that it's difficult if not impossible for us to agree on what is and isn't safe for work.
I'm aware of the issue, and trying to fix it. But medium won't let me make any edits. Since I originally published the blog have been in constant touch with their support but they all seem to be away for thanksgiving. Just keep getting the error "Oops! Something happened to Medium. We’ll fix it as soon as we can."
stevenicr your'e right in that regard. Yahoo is the only one who has provided an on premises solution which is really sad. Even they haven't released the dataset of their images just the model. If more of these companies released open source models on a frequent basis we would all have less objectionable content on the internet.
I guess the reason companies haven't done that and the reason Yahoo isn't really good is because it's difficult to constantly keep updating a model that's already been "released".
1) it's very difficult to find nsfw images especially a particular kind like gore or suggestive nudity unless you Google things (which indicates a bigger problem). Maybe the solution is to use Bing (maybe this would cause the same issue in compariosn) or DuckDuckGo. But honestly I think if DuckDuckGo indexed a page, I'm pretty sure Google did as well. You would probably need something off of non indexed website which makes the job significantly harder.
2) even though google has all the images it's still not the best performing NSFW Detector Nanonets is.
I wrote a comparison of all popular content moderation APIs on the Internet. My basic understanding from doing this is that it is far from a solved problem and there is no 1 size fits all.