Scooperland is a web-based image-sharing application whereby images are stored in shared Google Drive folders. The application can be customized to any desired degree using a scripting language called Cooperscript. Machine vision and machine learning are used to crawl the shared images and detect male/female models portrayed in those images. Various model attributes are detected such as gender, clothing being worn, hair color, orientation of body parts, etc. The application is written in Java and is bundled with an embedded open source web server called Jetty. The sister project of Scooperland is Cooprnet.com, which is a web hosting service for apps written in Cooperscript and Coopertags. Coopertags is a simplified version of HTML.
Business Model
Subscribers pay $18 per year or $2 per month if billed monthly. Subscribers can access all images shared by other users, and guests (users who are non-subscribers) can access all images shared by subscribers. Guests accessing images shared by other guests face a restriction: half of the images are pixelated versions, and half of the images are normal versions. Any user can email to any other user links to shared image folders, and all of those images are normal (not pixelated). Guests can make one or more image folders hidden (not shared), but all hidden folders reside in the same parent folder and all hidden folder names are numbered: 001, 002, 003, etc. Subscribers do not face those restrictions in regards to their hidden folders. Every image has a corresponding pixelated image having a width and height both equal to one-eighth of the width and height of the original image, respectively. The pixelated image file names begin with a dot (.) character.
Implementation Steps
- Implement Cooperscript 0.1, console-based
- Token parsing and building program tree has already been implemented
- Finish Cooperscript 1.0, console-based
- Make Cooperscript web-based
- Write main (default) Cooperscript script
- Research machine vision and machine learning
- Enable downloading of shared images
- Text search: captions/tags
- Make searching indexed
- Design website
- Launch website (no subscription fees)
- Beta test Scooperland
- Advertise using Google AdWords
- Make pitch to Ryerson DMZ tech incubator
- Hire Java programmer as Co-Founder: Ryerson recent CS grad
- Pay Co-Founder $4000/month + 50% equity
- Co-Founder to develop machine vision/machine learning aspects
- Mike to develop IDE in meantime, an Eclipse plug-in
- Search for angel investor
- After 6 months since co-founder hired, and no angel investor:
- Co-Founder is laid off
- Leave Ryerson DMZ
- Charge subscription fees
- Accept credit card payments
- Implement Cooprnet
Metadata
Each user has a corresponding SQL database stored in the root image folder. Every record in that database contains the following fields:
- id
- nextid
- parid
- childid
- hidden
- width
- height
- filename
- caption
- tags
The filename, caption and tags fields each contain up to 255 characters of text. Each tag in the tags field is blank-separated and contains only alphanumeric characters, hyphens, and underscores. By convention, a tag corresponding to the name of a model begins with a single underscore, and includes an optional underscore separating the given name from the surname. By default, image folders with a nsfw tag are hidden from the user. The hidden field (a yes/no flag) associated with an image folder record is false for shared image folders. The hidden flag is true if the associated image (downloaded by a guest from another guest) is pixelated.