Updates:
Facial Recognition: (Ryan, Kurtis, Brad)
On Thursday, we brought the mirror to the lecture classroom, where the Mother of All Demos (MOAD) will take place this upcoming Thursday. This was in an attempt to keep the lighting consistent in the room while we fill the magic mirror library with quality negative profiles. The most appropriate light setting on the light switch was determined as the second “Half” toggle switch. As a result, we hope that the facial recognition will be more consistently correct. We also toyed around the system so that more facial matches in a row would have to occur to identify a user.
Mobile Application: Updates (Brandon)
The customization feature for the Smart Mirror has been implemented. When the user clicks the “Save” button on the Web App the customizations of what the user wants to show or hide will be saved to their profile. The customizations for each profile will be pulled up when the registered user is recognized. If the user just registered, default settings will be pulled up until they personalize their settings.
Gesture Control (Ryan, Kurtis, Brad, Brandon)
As a group, it was decided that mounting a breadboard on the outside of the mirror would be less work-time-efficient than re-implementing the button in software. A HTML button module is linked from the web app on a new user’s registration to recreate the physical button hardware behavior.
The procedure for our mirror training module is now:
- If you are a user already stored in the database, nothing needs to happen (training already done).
- Else, to create an account, register with a new username and password at the displayed web address on a browser
- The HTML page with the log-in screen switches to a webpage with the software-defined button
- The button can be pressed to start the training protocol, which takes 10 pictures of the user in front of the mirror
- The training module completes and the mirror program restarts to read the newly created configuration file
Smart Mirror ON/OFF Protocol (Brad)
The mounted breadboard was decided against for aesthetic reasons and because we were having trouble deciding where on the mirror frame to mount it on. This means that the proximity sensor won’t be implemented in the final mirror build. As a result, we should be able to dedicate more focus on refining the facial training code to be as accurate as possible for the MOAD.
Smart Mirror Frame (Ryan, Kurtis, Brandon, Brad)
Some of the bubbles in the mirror frame stayed after time on the one side, and the other side is slightly scratched. We will have to finalize a decision for which mirror side is the worse of the two to peel off. The mirror frame can stand by itself and is made out of stained wood. This will house the acrylic frame and the monitor behind it so that our mirror can look aesthetically pleasing and be easily transported. We will also glue in the mirror and tv screen so that it is stable and will not fall out.
Next Steps:
Facial Recognition: (Ryan, Kurtis, Brad)
The only thing left to do is do as much testing before our final demo. Since stranger detection seems to be the hardest for us right now, we are trying to find volunteers to step in front of our mirror and see if they trigger a false positive. If so, we can play with thresholding and the consecutive number of times a face is recognized before logging in the user.
Mobile Application: (Brandon)
The Web App is fully integrated and is complete for the project.
Gesture Control: (Ryan, Kurtis, Brad)
The button is fully integrated as software and works to trigger the training system.
Smart Mirror ON/OFF Protocol (Brad)
The software definition of the protocol was implemented but the physical hardware still has to be tweaked. We decided to scrap a mounted breadboard and the proximity sensor, in favor of tailoring our efforts towards the main goal of the MOAD.