- Groups officially formed today
- Initial pitch of idea to other lab participants, Angelique Manella, and mentor Ethan
- Received feedback:
- Formulate timeline for deliverables starting backwards
- Determine minimum viable product
- Start testing at small levels
- Consider further training on top of trained models
- Juliette and Julie...
- ... sketched out timeline
- ... broke down project into smaller tasks
- ... completed Phase 1: Image Recognition API
- ... tried out two TensorFlow/Inception based app implementations on Android
- ... refined timeline and deliverables
- Listened to C2RO CEO presentation on product dev
- Canvased business model
- Updated repo README
- Created summary document to synthesize and track progress
- Tested 2 implementations on phones
- Deliverables of Phase 2:
- COMPLETE
- Wrote Java code for text layer in Android app to categorise items
- Successful trial run on mobile
- IN PROGRESS
- Improving trained model for image recognition
- Developing lists for recyclable and compostable objects for text categorisation layer
- COMPLETE
- No workshops today! Purely work kind of day
- Within improving the image recognition and classification,
- Went through process for re-training categories on TensorFlow demo example
- Figured out optimization of output .pb and .txt files to load on Android
- Identifying image sets to include in compostable list
- Tasks
- COMPLETE
- Added splash screen to the Android app and other UI changes
- IN PROGRESS
- Improving trained model for image recognition
- Juliette managed to train the model on our giant set of objects! Took 2.5 hrs
- COMPLETE
- St Jean Baptiste! Day off!
- Another day off since we're ahead of schedule
- All the buildings on campus were locked. Annoying.
- We finally made our way to Starbucks to work, thank god for Starbucks
- Where we're at:
- Have the newly trained model
- Works on TensorFlow for Poets Android App
- Continuous object recognition / detection rather than button push
- Does not work on our version... why? SAD!
- Remaining Steps:
- Make compatible with our version of app OR modify TensorFlow for Poets Android App
- Record demo video for presentation
- Make presentation
- Schedule
- Monday:
- Check in with Angelique
- Record demo of app with first trained
- JT: Orange, plastic waterbottle, paper towel
- Isabella: Tea bag
- Juliete: Bread
- Train another graph with more images
- Tuesday:
- Test new graph
- Investigate changes in confidence percentages
- Record demo of app
- Wednesday:
- Make / finish presentation
- Thursday:
- Practice, present!
- Monday:
- Check in with Angelique - went well
- Next steps:
- Isabella - cleaning category images => DONE
- Re-train graph and transfer to app => Juliette tonight to train graph
- Tomorrow:
- Follow June 27 schedule in previous post