hanacompsci

[hanacompsci]

update:

- will not implement generate solution 2 (maybe if more time) - view button, shows most recently saved list

[hanacompsci]

update:

change the user flow so that the list is saved only once the list has “passed”, at the results screen

[hanacompsci]

todo:

- fix myPackingList class (list not updated after solution selected) https://stackoverflow.com/questions/5545217/back-button-and-refreshing-previous-activity/47521055 - generateSolution2  - update database  - home activity/view button  - design and add logo

[hanacompsci]

generateSolution function does not work as it involves testing the solution on a copy of the list, instead alters the list

[hanacompsci]

update:

coded function to generate solution one; removes items with the most occurrences until condition is met and returns a list of items removed

[hanacompsci]

update:

change default sdk from 16 to 19 (min sdk required for tensorflow lite)

[hanacompsci]

update:

- original: dialog box confirm item > dialog box confirm list > click results > results page

- changing the user flow: dialog box confirm item > display list > click results > dialog box confirm list >results page

- purpose: user can read current list, so they know what is missing/included currently

[hanacompsci]

update: 

- narrowed down the no. of classes recognised by the ml model

to do:

- update the db w/ relevant classes and corresponding weights

[hanacompsci]

comments:

- use of customised trained teachable machine model restricts no. of items that can be classified by app  - expanded range of classes would require larger database to store the corresponding weight, infinite size 

[hanacompsci]

update: 

- collected training data for toiletries and electronics classes - retrain ml model w/ teachable machine  - tested ml model and exported

to do:

- integrate model into app - transfer learning clothing dataset 

[hanacompsci]

update:

- created new activity to navigate to - adds new item to list and displays in new activity  - 

to do:

- retrain ml teachable with toiletries and electronics classs - test and export model - transfer learning clothing dataset

[hanacompsci]

update: 

- app displays top three classifications in table - user can select correct classification from row in table (row turns grey) - created confirmation button; displays dialog box for user to confirm item and adds to list 'packingList' of 'checkList' objects 

 To do:

- navigate to new activity/fragment? - retrain default ML model to required classes (transfer learning, python, tensor flow)

[hanacompsci]

update:

- created a simple app using tensorflow image classification starter model  - app presents two options: select existing photo or take new photo - the app predicts the photographed item and displays top three classifications  to do:

- display top three classifications in list - enable user to select from list correct classification  - create class for confirmed packing list items, store item and weight

[hanacompsci]

created app activity: select photo from phone and pass through machine learning model, print highest result

attempt1: print top classification result  attempt2: print top two results, create a new function to find the next highest value, passing the index of the max that has already been found to exclude cons: does not work

[hanacompsci]

Update:

Completed the Machine Learning Crash Course: ML Concepts by Google Developers

[hanacompsci]

Update: 

Expanded the training data by increasing the number of classes (types of clothing) and data samples, images per class by doubling it from 20 to 40

(include results)

[hanacompsci]

Update:

Results from testing the model made from teachable machine showed some inaccuracies. Therefore, a need to retrain the model.

I plan to retrain the model by increasing the quantity of training data and adjusting the training variables; number of epochs etc