What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 133 Stack Exchange communities.

Is face detection/image comparison (a la Google Goggles) out of scope for Stack Overflow? This (albeit provocatively named) question was closed and moved (twice) off SO and I'm scratching my head as to why. What do the 5 users and 1 mod see that I don't?

share|improve this question
    
Link to the question in question: stackoverflow.com/questions/4998138/… –  Rebecca Chernoff Feb 15 '11 at 3:28
    
Check out the first answer posted to the question over at Stats.SE. Boy, was I overthinking it. –  Robert Harvey Feb 15 '11 at 3:39
    
It got another 16 or so upvotes at Stats.SE. LOL. Sure, we'll take your poor, unloved question. –  Robert Harvey Feb 15 '11 at 3:41
    
@RobertHarvey, I'm not as convinced as you are that that is such a great answer honestly. Sure, 5 people closed it, most likely because they saw Bieber and didn't actually read. That's what reopen votes are for. Doesn't mean a question is truly unloved. –  Rebecca Chernoff Feb 15 '11 at 3:51
    
@Rebecca: The question can get another bite at the apple on SO, if you feel strongly about it. –  Robert Harvey Feb 15 '11 at 3:55
1  
@RobertHarvey, my singular opinion isn't what matters. (: –  Rebecca Chernoff Feb 15 '11 at 4:17
    
@Rob: Most of the answers it's received at Stats.SE I think I understand well enough to know they're not stats related, but programming. –  Nick T Feb 15 '11 at 4:36

3 Answers 3

up vote 6 down vote accepted

The question isn't the best, but the migration to Cross Validated doesn't make sense to me.

Closing a question because it's unclear is ok. If it's a bad question, why migrate? Closed questions can be edited for clarity and then potentially reopened.

I don't see anything horribly wrong with the title either. It gets people attention, people just need to read. For me, the issue is that it says In order to filter them out, how can I tell programmatically whether a picture is that of Justin Bieber? without any indication of effort into accomplishing this.

Regardless, where are the comments asking for clarification? Where are the comments explaining why the post was closed, the guiding comments?

share|improve this answer
    
The only comment on the post that seems relevant to the closing was from 0xA3, who remarked, "Well, the way it was tagged obviously made it look like [a joke question]. –  Robert Harvey Feb 15 '11 at 4:20
2  
I suggested that it was overly broad due to the lack of detail just as indicated in your third paragraph. In the end, a Stats mod actually requested the migration, provided it received the non-exclusive touch-up that it did get. To quote the mod, "This is a plain image classification problem, so good old Machine Learning which is just in scope of CrossValidated. Stats.SE is a deceiving name." Doesn't explain why it doesn't belong on SO, but it does explain that they did want it on CrossValidated. –  Grace Note Feb 15 '11 at 13:26

I think the op needs to make certain that the question is firmly in the area of programming.

Simply saying, "how do I X, but programmatically?" isn't going to get a good response, especially if it shows the op has done little work. Essentially it becomes a plzsendmethecodez question.

If this question is going to pass muster, it's going to have to be simplified and broken down into smaller parts.

"how do I do image recognition?" would be a good start, for instance. Alternately "how do I find a particular person's face in a set of images."

share|improve this answer
    
I figured the question was asking for a high-level "how do I go about this", it didn't seem like a blatant plzsendtehcodez query. –  Nick T Feb 15 '11 at 4:00
    
"A problem I've encountered is that after I filter out all but the images that are clear portrait photos..." suggests that the OP is already using some sort of software to do image analyais and face recognition. Then they essentially ask how do they perform facial identification for a particular face. They don't indicate how they filtered out all images except faces in the first place, they don't specify the software they are using, or any details that suggest they have actually completed what they say they have, or that they are capable of going to the next step without comprehensive help. –  Adam Davis Feb 15 '11 at 5:53
    
@Nick Honestly it appear that they are doing some sort of statistical analysis using some packaged software. Yet the OPs question and answer history suggests that they understand programming. So why would they indicate that they are able to do something relatively complex, but not go the next step without mentioning how they are doing the first (and absolutely related) part? –  Adam Davis Feb 15 '11 at 5:57
    
As far as I can tell, it's valentines day trolling on stack overflow. Why it was migrated instead of closed I don't know, but the folks on stats have given him about as much help as he could have expected on SO, so he's not worse off by it. –  Adam Davis Feb 15 '11 at 5:58

For one thing, that question is arguably a duplicate of the question Face Recognition Logic, so it could be closed on those grounds as well.

I'd say that there certainly is a place for questions about face recognition or image processing, but those questions need to be much more specific than this one. Face recognition is a huge problem that many people have spent a lot of time working on, so you need to whittle down to a specific aspect of the problem.

For example, it's perfectly legitimate to ask if there are any existing face recognition libraries for the iPhone, or how to integrate OpenCV with an Android project. These are questions where you've identified a platform and a specific task you're trying to perform, so people can give clear answers that won't turn into research survey papers.

Other aspects of these problems, such as specific algorithms, math, or even the psychology of image analysis may be appropriate for other sites in the network, but if someone has a targeted code-related question, it's more than welcome on Stack Overflow.

share|improve this answer
    
Face recognition is a many to many problem, in most cases. The OP's question was about one to many recognition, a much simpler problem. They have the same issue with fingerprinting systems; bioinformatic systems that allow a small group of users access to a particular area are much easier to design than, say, search systems for crime fingerprint identification, because the domain space is much smaller. –  Robert Harvey Feb 15 '11 at 3:57
    
@Robert - True, this question has a narrower scope than others I've seen on the topic, but I'd still like to see it made even more specific. Even this particular aspect can be tackled in a number of ways, so providing a platform of choice, or language, or some other specification would really help to make the question more approachable. –  Brad Larson Feb 15 '11 at 4:19

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .