Week
|
Date
|
Scheduled Features
|
Completed Features
|
Pending Features/Remarks
|
1
|
17/08/12 – 24/08/12
|
- Explore Classifier Algorithm (CA)
- Measurement collection
|
- Theoretical understanding of Neural Network (NN) and its classification feature
- Measurement collection
|
N.A.
|
2
|
25/08/12 – 31/08/12
|
- Analysis of measurement collection
- Gender differences trend establishment
- Gender recognition based on Waist-to-Hip Ratio (WHR)
- NN classification
|
- Analysis of physical measurement collection classification feature
- Gender differences trend establishment based on physical measurement analysis & secondary research
- Measurement of height based on image captured by Kinect
|
- Found that waist cannot be detected by Kinect, hence dropping WHR
- Difficulties encountered in implementing NN classification
|
3
|
01/09/12 – 08/09/12
|
- NN classification & scoring system
- Parameters expansion: neck & shoulder width
|
- Passing of height and shoulder width captured by Kinect to NN
- NN classification based on height and shoulder width
|
- Decided to include parameters hasBag & hasSkirt instead of neck width in the upcoming week
|
4
|
09/09/12 – 16/09/12
|
- Gender recognition based on complete set of inputs
- Capturing shopper’s outline
- Countdown timer
|
- Gender recognition based on height & shoulder width
- hasBag & hasSkirt displayed for consideration
- Countdown timer
- Refining of shopper's outline
|
- hasSkirt, hasBag parameters to be improved and integrated with passing in of parameters through Kinect
|
5
|
17/09/12 – 23/09/12
|
- Capturing of photo
- Backward propagation
- Saving learning state
- Integration of photo taking with gender recognition
|
- Capturing of photo
- Backward propagation
- Integration of photo taking with gender recognition
|
- Saving learning state is pushed back to Iteration 6
|
6
|
24/09/12 – 29/09/12
|
- Microsoft Tag
- Advertisement Management System page
- Refining overlaying of augmented background
- Integration of Microsoft Tag into photos
- Addition of behavioural parameters for gender recognition
- Integration of gender recognition (with behavioural parameters) with photo taking
|
- Microsoft Tag
- Advertisement Management System page
- Refining overlaying of augmented background
- Integration of Microsoft Tag into photos
|
- Addition of behavioural parameters for gender recognition
- Integration of gender recognition (with behavioural parameters) with photo taking
Behavioural parameters were not very accurate
|
7
|
01/10/12 – 06/10/12
|
- Using real figures instead of Kinect figures
- Addition of behavioural parameters for gender recognition
- Integration of gender recognition (with behavioural parameters) with photo taking
- Integration of Microsoft Tag, gender recognition, and photo taking
|
- Using real figures instead of Kinect figures
- Addition of behavioural parameters for gender recognition
- Integration of gender recognition (with behavioural parameters) with photo taking
- Integration of Microsoft Tag, gender recognition, and photo taking
|
N.A.
|
8
|
08/10/12 – 12/10/12
|
- Scene transition
- Display shopper's silhouette
- Make callout bubble dynamic
|
- Code clean up
- Scene transition
- Display shopper's silhouette
- Integration of outline-photo-taking-displaying photo with tag
|
- Make callout bubble dynamic
|
9
|
15/10/12 – 19/10/12
|
- Choosing of door
- Unsupervised learning
- Make callout bubble dynamic
- Stick man approaching chosen door
- Debugging of bug: crashing in the wild
|
- Choosing of door
- Auto-rescale callout bubble
- Stick man approaching chosen door
- Recording of learning response to database
- Debugging of head-depth bug causing crash in wild environment
|
- Callout bubble position needs to be relative to user's
- Integration of unsupervised learning to AlterSense
|
10
|
22/10/12 – 03/11/12
|
- Relating door scene image with photo taking scene
- Integration of unsupervised learning to AlterSense
- Callout bubble to follow (relative to)user's position
- Activation of training upon detection of dropping accuracy
|
|
|