No level content filter in Super Mario Maker, but obscene content can be reported

- will not have a level filter for rude/obscene/distasteful content
- any level created can be uploaded and downloaded straight away, so long as it can be completed
- there will be a “report” button for users to click if they find content in a level to be inappropriate content-wise
- Nintendo will review the report and make a decision after that process

Thanks to nn64 for the heads up!

Categories: Consoles
Tags: eshop, wii-u, mario
Games: Mario Maker


How could there even be a filter for obscene content? I'm not sure the game could register you're making the blocks in the form of a dick or anything like that.

Thu Jul 02 15 07:49am
(Updated 1 time)

morphology recognition algorithm work with things that look like the thing they were trained to recognice. For example, if you train a haar cascade to recognice faces and you draw a face it will probably will recognice it as a face, even if it is only eyes, eyebrows, nose and mouth. Even if you say... that cloud look like a face, a poorly trained algorithm will also thing that the cloud is a face if you didnt point out that the cloud is a negative in the trainig set. All you need is an adecuate database to train your detector and it probably will detect even minecraft like penises.
Also, the game doesnt need to recognice them, the server where you upload the level should do that work.

It would probably lead to a lot of things recognized as something obscene when it's really not. Especially since the entire game is tile based. You only need a vague rectangle with two circles for something to look like a dick, for example.

I don't think it would be worth the trouble creating an adequate enough filter for this. Other Nintendo services, like Miiverse for example, haven't had any major issues with this kind of content either.

the biggest problem would be creating a database of "blocky penises". the rest of the process is pretty much standard and automatized. But you are right, the ammount of work needed to put a database big enough to reduce the false positives to a reasonable number would be of a few thousands of "positive images" and 10 times that for negative images. if they use lbp,hog and svm training the result would be good enought (my guess is 93% hitrate with 0% of false positive or 98% hitrate with 3% of false positive)

Yeah, makes sense. It's too hard to register something like that.

Thu Jul 02 15 03:11am
(Updated 1 time)

Well, duh, that would be incredibly complicated.

I can see it already, dicks, dicks everywhere!

oh man I'm going to spend hours reporting content, it sucks but that's online communities for you

Want to join this discussion?

You should like, totally log in or sign up!