TODO

TODO: add pollution event visualization
  - show a timeline with events detected by the deep learning model
  - show videos of notable events detected by the deep learning model

TODO: improve security
  - if the user make 6 wrong batches in a row, mark it as an unreliable user for one day?
  - if a video jwt is rejected due to insufficient nbf (impossible for normal users), mark the client as a spammer
  - detect and mark spamming (e.g., all not selected, all selected) in the database

TODO: add active learning mode for labeling
  - implement a mode that the batches are selected from a deep learning model

TODO: design features for sharing achievements
  - add a leaderboard for showing user id and scores
  - generate PDF certificates of label statistics for each user upon request
  - add open badges (complete tutorial, label X effective batches, keep labeling for X days)
  - allow users to share the badge with their achievement on social media

TODO: add the feature for labeling smoke opacity
  - add a compass to show sun direction for each video?

TODO: add the feature for selecting the shape and size of the smoke emission
  - multiple options with graphs (of various shape and size)?

TODO: refactor code based on https://codeburst.io/jwt-authorization-in-flask-c63c1acf4eeb

TODO: as users gain enough scores, advance them to the harder mode
  - laypeople mode: select videos that have smoke
  - amateur mode: selecting the shape and size of the smoke emission
  - expert mode: label smoke opacity (low, medium, high)
  - how to invalidate previous user tokens with different permissions?
  - need to add a table to record the promotion history
  - need to encode client type in the user token, and check if this matches the database record
  - for a user that did not login via google, always treat them as laypeople

TODO: improve the tutorial
  - add the "back" button to go back to a previous step

BUG: when changing the client type of a user, previous tokens with different permissions are still working
  - need to invalidate previous issued tokens?

OPTIONAL: graphically display the label statistics (instead of using text)
OPTIONAL: add a playback timeline bar to show the video playback time
OPTIONAL: add a link back to time machine viewer (or a larger video) on the labeling page (also gallery page)
OPTIONAL: prevent the case that multiple people label the same data
  - add the last_queried_time to video and query the ones with last_queried_time <= current_time - lock_time
OPTIONAL: design a sticker at the side to remind people the three rules to label smoke
OPTIONAL: think about how to engage online audience for smoke labeling