Sprint #6 Retrospective

[qspraut]

Sprint #6 Retrospective:

• members in attendance:
  • [x] Alex
  • [x] Anand
  • [x] Brandon
  • [x] Quinn
• what worked well
  • iOS Image uploading error was resolved!
    • switched upload method when using app from a phone vs. a computer
  • We successfully implemented the confidence score on classification predictions
    • 
  • We experimented with new methods to determine calories from macros, and realized our previous method was more accurate
  • We fixed an error with cv2 color detection on certain images due to incorrect file paths
  • We updated our model with more training images and better volume estimation
  • Our classification is deployed to Heroku and working properly
  • App is able to display Spoonacular recipes
  • Some style revamping
  • Added 'simple' form completion - allows users to skip logging things like custom name and mood if they dont feel like it
    • 
  • Added ability to delete foods from the user history page
    • ignore my face, did not have food with me when taking screenshots
    • 
• what didn't
  • Had trouble integrating spoonacular calls upon rendering on the expanded meal cards
  • We tried experimenting with image segmentation but were unable to come up with an accurate enough function
  • Some issues with processing spoonacular and then displaying them
  • Some images are still being incorrectly classified. We are working on improvements to our model to improve accuracy
  • We are struggling to find a balance between improved accuracy and overtraining on certain foods
  • We downloaded a set of new training images that ultimately failed; they were too artificial (were more models of foods than actual foods on plates)

• self assessment on progress

  • where are you in relation to progress towards product and milestones?

  • Approximately 87% done overall (75% done with progress for this term)

  • We have many stretch features like the food image community and confidence score on classifications implemented, and are working on further polishing the features

  • lay out each of the following weeks till end of term with brief goals for each

    • Current week: Finish stretch features for food recommendations based on goals and history, improve model accuracy with more training images and iterate through validation process, flesh out recipe suggestions, flesh out food journaling
    • week 8: Write medium article and market copy to tech community
    • week 9: Make final adjustments and test with more users
• briefly summarize any other topics/discussions
  • This week was very big for fixing large errors like issues with iPhone deployment and confidence scores on classifications. We are going to carry this momentum into next week to finalize a product that we will all be happy with.

[timofei7]

yay, glad to hear about all the big updates!