3. For people who are on webex Please visit this URL for all the videos that we play. http://tinyurl.com/twdemo
4. We built “ An augmented reality application to be used inside an Enterprise to manage contacts and location services. “
5. Augmented Reality? Augmented reality (AR) is a term for a live direct or an indirect view of a physical, real-world environment whose elements are augmented by computer-generated sensory input, such as sound or graphics Simple Defn : overlaying of digital data on the real world. * Courtesy – wikipedia and howstuffworks.com
18. Step 1 Detection /Scanning No faces in this image Face Detection –algorithm to detect if there is a face present on the given image.
19. Step 2 Recognition Point out the faces that were detected Face Recognition – Once a face is detected, the face has to be processed by an algorithm which will associate a unique id from the training set.
20. Step 3 Data/Action Fetch Augement real time image with info bubble PrateekshaUday nono@hotmail.com 9889090123 Once the algorithm returns a unique identifier based on the processed image, we fetch the data corresponding to that id. In our case, the details from Active Directory. (actually, not just AD)
24. First Approach Training Set Manual Training Face Detection LDAP Face Recognition Service Calls App Massive Database of Photos
25. Problems with this approach Too Painful Training Set Manual Training Face Detection Face Recognition Service Calls App Not Real time – Not scalable
26. We found Open Source Computer Vision. A free library that does object detection, recognition and 500 other image processing algorithms. Yay! OpenCV is a C++ library
27. Second Approach Gives Compile Using opencv.so C++ Libraries Android NDK opencv.so Java Native Interface (JNI) App
28. Pros & Cons - Second Approach opencv.so Java Native Interface (JNI) App Not Accurate Near Real time No Service Calls Poor Training data Totally in Phone
29. Hybrid Approach Face Detection Scanning Mode Face? Yes Face Recognition Training Mode
30. A new way of putting up Assistance and event posters. Posters
31. Contacts Problems Wait! Is there a name of a person as well? So, now I know the person as well his/her phone number. What happens if the Phone number changes? Exposed Phone numbers are dangerous. Can be abused by telemarketers
32. A new way of putting up assistance posters. No numbers, no hassles. For all Travel related Queries
33. X-Conf 4 – March 5th Event Posters - Revisited //TODO: Add functionality that will add this event to your calendar and notify you 10 mins before the event Title: Don't do that, then! (Doctor, it hurts when I rewrite legacy applications) * Photo courtesy thoughtworker.com
34. Presence * VISION @starkcoffee - hopefully my.thoughtworks will be a completely fresh experience to TWers !
35. * Almost there Dinesh Tantri – Geek Lunch Talk: Introducing my.thoughtworks Give Feedback , ask questions – use panacea Event / Conference Feedback – 1
36. Event / Conference Feedback - 2 * Almost there Talk: Introducing my.thoughtworks! Comments Feedback
37. * VISION Visual URL’s and Bookmarking Use OCR to save this URL on a poster or on a hoarding.
38. Wait, does this mean we need to train photos of every employee?Damn, that’s painful.Luckily . . . We have a solution
42. Do you know these applications? * Logos courtesy their corresponding websites
43. Location Problems Only 13% of the smart phones sold across the world are Location Aware Less than 4% of all mobile phones sold last year were Location aware. * Stats courtesy: http://textopiablog.wordpress.com/2010/02/22/how-many-people-own-a-location-aware-device/
44. Location Problems Too many places. Search doesn’t solve the problem either. *screen shot courtesy google.com
45. Location Problems Roof top Restaurant Roof top - Bar Pizzeria Chinese Restaurant Cafe Multiple Avenues – Same GPS Location
47. But, how? Most smartphones come with a good camera. Given that we already had the technology to recognize faces, we just appropriated it to make a location service just like FourSquare or Gowalla. (our algorithms can be trained on other objects as well, like boards, toys etc)
53. Mockups * We used http://iphonemockup.lkmc.ch/ for smart phone mockups
54. Application Architecture Main UI Surface Overlays Camera Info Overlay status Overlay Processing Android native detection Recognition service Gab server Location server PANACEA