Unity3d and python implementing an interface between a unity scene and mechanical turk

Final Goal: Develop an interface between Amazon Mechanical Turk and Unity in which turkers are asked to find objects in virtual scenes and we record the paths they take for doing so. DEMO: This demo will show you what has been done already. The installation instructions are: On linux run this lines to install the program. You need python3 for it. On a terminal: pip install [login to view URL] wget [login to view URL] wget [login to view URL] wget [login to view URL] unzip [login to view URL] unzip [login to view URL] NOTE: on MACOSX download the osx version instead of the linux version. wget ---- Then run the script example. On a terminal: python3 [login to view URL] Wait till the game loads and then use the keys to move around: w,a,s,d: usual moveahead, back.. q,e: rotate left,right respectively r,f: look up and down respectively You can access the target image directory that is being printed in the terminal to see the image of the object you must find. When you find the object function target_found() returns true. The event contains metadata, one part of it is 'lastActionSuccess' which is true when the character was not blocked by an object when trying to move. What needs to be done in the software department: I have the scenes to be edited in unity (we used unity 5.4.0f3) so we could actually implement everything in unity (csharp) to deploy in webgl for mechanical turk. Or we could ask the turkers to download the linux/osx scenes and play on their laptops. In any way this should be the final functionality of the program: • One turker can be placed in one scene only once (the idea is that we collect paths of turkers that didn't know the scene beforehand) • The turkers are aware of the object they need to find in the scene they are • The turkers must be able to navigate in order to find the object they are intended to look for • The program should record the starting position, the paths taken (sequence of actions, possibly with metadata) and save them in a database. I was thinking about using firebase for this thing • The turkers must be paid when they find the object (target_found returns true) Other deliverables: • Documentation that exhaustively explain the functionality of the program, including screenshots, examples, a recorded video that explains what modules act in what situations, ETC. I should be able to replicate the results easily by looking at the documentation. • Presentation: I will schedule a presentation with you so that you show me the development of the project • Sandboxed version of mechanical turk: I should see a sandboxed version of mechanical turk that works and has full functionality • Final version of mechanical turk. Let's discuss further details when we chat.