Well, this is insanely creepy, but at least it shows a bit of entrepreneurial spirit. Rather than focus on the somewhat questionable ethics, how about we talk about how it actually accomplishes this?
1) What features of OpenCV are you using, what sort of shapes are you looking for in images?
2) What's the true positive rate on test sets? And importantly, what's the false negative rate?
1) We are using rapid Object Detection by training Haar classifiers of different bikini shapes (top, bottom, strapless bikinis etc). We used opencv_haartraining to get a few XMLs we use for testing each image.
2) The true positive and false positive rates vary depending on the test we are performing (top, bottom...) so we have a formula based on the combination of these and the amount of hits found in each picture by each classifier, where we assign the picture a probability of it containing a bikini (based on a controlled sample of 3000 pictures). As of now we are looking at user feedback to make sure our threshold is where it needs to be so that we provide just the right amount of true/false positives in our results.
As the poster has already pointed out, these are pics that people have already shared via facebook.
Computers sometimes allow us to do things was were already doing, except at a massive scale. Pre-napster, people were already able to share music with their friends via mix tapes, etc. Suddenly they were able to share with millions, without leaving their own home.
Similarly people were already able to scroll through their friends facebook albums for bikini pics. Suddenly they are able to to do it for 5 friends at once, with the click of a button. Changing the difficulty and scale of something can totally alter the dynamics of the situation. It will be interesting to see if services like this alter the behaviour of people 'oversharing' on facebook.
Well I'm hoping I'm not the only user, and yes all my friends know I'm working on this. And have been for a while. It's not about perving your friends' pictures or turning people into sex objects. These pictures are already up on Facebook, Badabing is just a filter. We will be adding a feature to nominate your friends to a contest, as long as they accept to be in it.
You are only filtering content already available on Facebook, there are many apps that do something similar. As far as the black bar on the screenshots, I actually got everyone on the preview pictures to agree to be on the App Store and website, they are all people I know and actual results from a search using the app.
Thanks for the feedback. I respectfully disagree and hope that is not the case. In short, the app is an image recognition search engine for social networks. Beach and pool pics is the first "search keyword" we are using since we assume it is one of the more popular, but there are more to come.
1) What features of OpenCV are you using, what sort of shapes are you looking for in images?
2) What's the true positive rate on test sets? And importantly, what's the false negative rate?