We already have a large collection of d1ck pics, it’s the going through them and cropping them down to the offending item is what I’m trying to avoid. I need my eyes!
From: [email protected] [mailto:[email protected]] On Behalf Of David Connors Sent: Thursday, 4 February 2016 12:58 p.m. To: ozDotNet <[email protected]> Subject: Re: Haar Cascade for image classification. I can't help you out in your quest for a corpus of d1ck pics but I did notice in passing the other day that Google's cloud vision API identifies naughty bits if you ask it to. https://cloud.google.com/vision/ On Thu, 4 Feb 2016 at 12:51 David Smith <[email protected]<mailto:[email protected]>> wrote: Hi all, We run a dating website and make use of Emgu & OpenCV to identify "face" pics of our members as they upload images and suggest a zoom and crop based on the result of the detection. This is working really great, however we would now like to extend our checking of pictures and perform a check for "adult" content. To do this I need another suitably trained cascade file. After reading the docs, this is going to be a time consuming and somewhat unpleasant process going through images to build a training set. The adult content we want to identify and flag are pictures containing a penis. My question and hope is, has any one already done this and able to share? Or has someone come across any pre-trained cascades online that we can plug in? David -- David Connors [email protected]<mailto:[email protected]> | @davidconnors | LinkedIn | +61 417 189 363
