Nudity And Skin Tone Detection Software Development (Type B)

Cancelled

Job Description

We are looking for a developer to create a Skin Tone and Nudity detection application, which will allow a user to select an individual file, or directory of files (including subdirectories), containing images and videos. The application will analyse each file and reveal the level of Skin Tone present, along with the amount of Nudity detection based upon the techniques listed below. The application will be controlled by a simple GUI (see details below), but also be able to be controlled via a API allowing commands to be issued from the command line, or for the application to be integrated into a larger Java application (yet to be developed).

The Nudity techniques to be implemented in this project are to be suggested and chosen by the developer. In your application you must state which techniques and how many you propose to implement, (more the better), how they work, along with any references or publication on the techniques. The chosen techniques must be as accurate as possible, with little or no dependencies (i.e. no need for training or a database etc).

*** There are two distinct version of this project, please feel free to bid on both. The differences between them are the Nudity recognition techniques, which are required to be implemented. Please also say if you could implement all techniques, as this will be very advantageous ***

Specification (deliverables) of the application tool to be developed:

1. A Java tool (or front end) that delivers the following functionality:

a) Determine the amount of Skin Tone visible in the image / video
• Including the ability for the user to turn on / off ‘Pantone colour’ detection to find various extended skin colours. Pantone colour reference included below.
• Detect faces in the image / video, and report the proportion amount of skin visible on the face against that on the rest of the image (this will help compensate for close up / far away images)
• Reporting results:
• ‘Low amounts of Skin Tone’ for 0-35%
• ‘Medium amounts of Skin Tone’ for 35-70%
• ‘High amounts of Skin tone’ for 70%+
• (Note: These results are based upon the proportion of skin in the image/video vs. the amount on the face, unless no face is detected, to which it is the overall skin tone visible in the image/video. The results should also state which is used e.g. ‘Overall Skin in multimedia’ or ‘Overall Skin vs. Facial Skin Amount’)
• There must be a setting to allow the user to change the percentage amount for each level of skin tone (i.e. to change low to 0-25%), and to turn facial detection on or off. Default values are above.

b) Determine the level of Nudity visible in the image / video, using:
• Techniques chosen by the developer

2. A log file must be produced showing the applications analysis detail, including:
a) Time and date of scan
b) Type of scan (directory or individual file including still image or video)
c) Settings of scan
d) Details of file or directory scanned
e) Overall scan result:
• Number of files scanned
• Types of files scanned
• Number with Low, Medium, High Skin Tone
• Number with Low, Medium, High levels of Nudity
• Files skipped (due to errors or which are unsupported)
f) Result for each file including:
• Amount of Skin tone
• Number of faces detected
• Proportion of Skin tone in whole image vs. Proportion of Skin tone on detected faces
• Nudity detection result
• Applicable confidence levels
• For Video, all these details must be included along with a breakdown of these results for every 1-second of the video, and an overall of what percentage of the video had high, medium and low nudity and skin tone. Example log for each second:
Time: 00:01:00, Skin tone: 50% Medium, Nudity: Medium

3. A simple GUI (based on the designs issued to shortlisted candidates) to run and control the application. To allow for quoting the GUI features:
a) Image and Video handling (playback, opening, thumbnail scrolling)
b) Simple checkboxes, buttons, colour coded labels, scrollbars, drop down boxes and a progress bar
c) Buttons to open selected files natively on OS, open external files e.g. PDF’s
d) Controlled by a simple settings option screen
e) 1 Main Application Window, and 1 Option Screen

Useful references:

• Pantone Colours: http://www.erdemyilmaz.com.tr/wp-content/uploads/Pantone_SkinTones.htm
• Nudity Detection: http://onebit.us/x/i/814381733331796005.pdf
• Nudity Detection based on Image Zoning (Google Search)
• Nudity Detection based on Bag of Visual features: http://dl.acm.org/citation.cfm?id=1730317
• Nudity Detection overview: http://publications.lib.chalmers.se/records/fulltext/76159.pdf


PLEASE SEE THE ATTACHED PDF DOCUMENT AND APPLICATION REQUIREMENTS BEFORE SUBMITTING - APPLICATIONS WHICH DO NOT PROVIDE THE INFORMATION REQUIRED WILL NOT BE ABLE TO BE CONSIDERED.

Open Attachment