Todo
Next: On branch cleanup
cleaning the code for the main app, next is to make the data downloading more robust.
- [ ] Create a new branch to keep cleaning up
- [ ] Data ideas
- Food images = Food5k + Food101 (random subset)
- Non-food images = Food5k + Open Images (random subset of 10000 images from training set)
- [ ] Clean up data download
- [ ] Use Open Images for non-food images:
- [ ] Use Open Images + Food101 for food images:
- Can also filter on Open Image for food images?
- Download random set of images from Open Images — filter food & not food
- Use Food101 random subset to create images of Food + could also download images from the internet of bulk food classes to classify food
- [ ] Clean up model training to reflect changes on data
- [ ] Make model training scriptable
- [ ] Clean up GitHub repo to make the app reproducible
- [ ] README
- [ ] Turn notebooks into scripts
- [ ] Fix prediction probablity for foodnotfood.app
Rules
- Every hour:
- 10 pushups, 10 squats, 10 kicks, 10 punches
- Surprises at 77-minutes in
Hours
- [x] 1
- [x] 2
- [x] 3
- [x] 4
- [x] 5
- [x] 6
- [x] 7
- [x] 8
- [x] 9