Nudity And Skin Tone Detection Software Development [Updated, Reduced Requirements]

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

1. A Java tool (or front end) that delivers the following functionality:
o Determine whether an image/video contains skin tone and whether is it possible to say that the skin amount and position of the determined skin tone equate to an image/video containing human nudity.

o Analyse, separate and move images/videos from an original source directory to a new set of multiple directories, named according to the level of determined nudity or amounts of skin tone that is determined to be present in the image (or not). Parameters are:

• No nudity 0-20% skin tone,
• Some nudity 21-60% skin tone,
• Lots of Nudity 61%+

2. Produce a log file detailing activity, including image file name, the category that has been determined and the directory it has been placed into. This log must also include the percentage amount of skin tone and nudity found in each image. For video, the log must report on the amount of skin tone and nudity through out (i.e. 00:05 seconds, 51% skin tone, nudity …)

3. A simple GUI to run and control the application, with the results displayed in tabs, 1 tab been an ‘overall result’ status of the scan (e.g. 100 files scanned, 10 no nudity, 80 nudity … etc), and 3 other tabs representing the parameters above (no nudity tab, some nudity tab, lots of nudity tab). In these tabs, should be thumbnails or small previews of the images/video, which can be clicked to load in the native viewer, or enable the user to manually change the result, e.g. change a image from no nudity to nudity.


Important accuracy related requirements:

• The methodology used to determine the amount of skin present in the image and whether or not that equates to ‘nudity’ or not, should consider accounting for when image / video is taken from a distance (e.g. whether the picture is taken of a person very close up, where a lot of skin colours would be present, or where the picture is taken of a person far away).

It has been suggested this can be done (but feel free to use any other methods you know) through:
- Trying to identify faces and proportioning the skin found on the face to that found on the rest of the image / video
- Body overlay or skin positioning, where the location of the skin found can be used to determine signs of nudity

• The application must compensate for different Skin colours. Our research shows other researchers have attempted to solve this issue by:
- Using and accounting for pantone Colours: http://www.erdemyilmaz.com.tr/wp-content/uploads/Pantone_SkinTones.htm
- Detecting faces, and identifying the skin colour on the face, and using this as the skin colour to look for in the rest of the image (however, beware instances where more than one person are featured in the image / video)


*** BEFORE APPLYING PLEASE NOTE: When sending proposals or messages, please include a statement confirming that you can deliver all of the above specifications and requirements, how you will develop the application and the estimated time it will take. This is a fixed price project, so please ensure your bid reflects this. No payments will be made up-front, but can be deposited on the system. The money will be released upon project completion ***


Technical detail:

- A desktop tool, preferably (but not necessarily) all written in Java as apposed to just the front-end
- Must be able to be combined and further developed and distributed as part of a larger Java application
- Must be provided to HurstTrentham with an Application Programming Interface
- All source code and functions will be plain text and commented
- A basic guide on how to run the application must be included, along with sample files for testing
- Must work on Windows, Mac and Linux
- Disclosure of which Skin tone / Nudity / facial analysis technique(s) have been used
- Any third-party code must be highlighted, and must be free for commercial usage by HurstTrentham
- Controlled fully by a basic simple to use graphical user front-end
- All source materials, including a number of actual tests must be included, along with a simple user guide on how to run, and what to expect
- The system must create a log file for each scanned image/video, which will store the workings and full results of the application.
- Must create a fully detailed overall log file for all files scanned
- The application must allow easy changing of thresholds for determining nudity and the groups to separate skin amounts into
- Please assume that on average images to be inputted to the system will be 1-2mb in size, and videos 100MB but may vary, in formats JPEG, PNG, BMP, AVI, MPG, MPEG, WMV, MKV and as many others as possible)
- The application should have the ability analyse one file at a time, or a directory of files at once
- The freelancer must collect test data, none will be provided