CS6825: Computer Vision word cloud
Instructor Outline Syllabus Projects Links  

ExtraCredit for Project 3's OPTION 1 - Kinect CPS Deployment Beta testing with experimentation

                              Proposal Due Nov. 15,
                             Project DueDec. 2 ,
                               (points 130)

Evaluation Guidelines 

This is extra credit and not required. You can choose this only if for Project 3 you decided to do Option 1- Kinect CPS

80 points You are going to actually deploy to a minimum of 3 users your application and perform experimentation and collection of real data (must be built into the application and uploaded to your server). Note the users must posses a Kinect system (so this will most likely be friends and family ---but, check out the local meetup on kinect developers for someone who might consider letting you install the Executable (don't give them your source code IF YOU CAN HELP IT --- you may have to install the source code and development environment :< ) on their computer and connect their connect up to this computer and let you as necessary configure the application (for registration of the RGB with depth data).

 

20 points If you are able to get your an overall approval rating of 6 out of 10 from your user base you get a full 20 points.

 

Deployment on min 3 Users Hardware

You will find 3 people that own Kinect sensors (you can not give them ours) who have a computer and are willing to connect their Kinect up to it and install your EXE application. Don't give out the source code. One issues is they may not have the power/usb connector necessary and this is a purchase you may need to offer them (see ebay/amazon for less than $10 for this item). You will need to minimally:

  • Save and Collect raw 10 section video for each time used when the user launches the app, you should at least during it first use from a new launch, capture a sequence of say 10 seconds of video after the first 1 mintue of running and send to your server account along with some unique user ID, datestamp and some launch id(unique to represent this user + launch). Because this is a desktop application, I will allow you to alternatively save the videos (given name that tells timestamp) on the computer.
  • Save and Collect results after the a new launch, the user takes some raw image and you will process it/them. Store the results of an image frame whenever a fallen user is detected (again you may save locally on the computer). You will need to revisit (or have user email you) the user to collect the raw video data and any detected user images.

Analysis, Improvements Needed, Report

You are going to create a report showing for the 3 users all the data you collected but, you are also going to interpret this data. Show me examples of it working and not ---give a tally of number of times it worked, number times it didn't, any discrepencies where you thought did work and user didn't or vice versa? show me the numeric results of the user feedback for each focused question.

You need to have each user fill out a say 10 question survey aboiut their opinion of how this system worked --- as part of it you must ask for a numeric 1(bad) to 10(excellent) overall rating of the system by the user.

 

You are going to give a DETAILED suggestion/plan on how you can use this information to improve your product before a real launch to the marketplace. Give algorithms where necessary. If everyone was really happy with your app--then instead you must give me what your 2.0 version of your app would do and the algorithm behind it.

 

PROGRAMMING (for 30 points of this extra credit) Find one thing you need to fix or if nothing a small future improvment (version 1.1) and program it (keep the version 1.0 stored separately in case you need to back out to it).
© Lynne Grewe