Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Swipe to accept/reject frigate+ detections #16143

Open
rjfehr opened this issue Jan 26, 2025 · 0 comments
Open

Swipe to accept/reject frigate+ detections #16143

rjfehr opened this issue Jan 26, 2025 · 0 comments
Labels
enhancement New feature or request

Comments

@rjfehr
Copy link

rjfehr commented Jan 26, 2025

I want to provide 300% more human feedback data on frigate+ detections, to accelerate AI model development, through a simplified UX that provides greater biochemical reward for my effort.

On the mobile landing screen for detections, frigate+ users can use common left/right swipe gestures to reject/confirm the detections without tapping on the image to reveal the existing yes/no buttons. Many possible alternative interactions to achieve this goal may exist within the mobile screen as described, and certainly in other form factors where gesture is not well suited.

TL;DR - More users will provide vastly more useful data to inform Frigate+ model development because…

  1. Less effort/complexity = more users completing more training tasks;
  2. Faster path to reward = neurochemical reward increases desire to engage;
  3. Leading to more active users, providing more data per user, as they complete tasks where humans excel and current AI models perform poorly.
  • Usability of this feature would be benefit from increasing the prominence of the bounding box and related details. Cropping the image and/or emphasizing the text elements for example would make the detailed image view less frequently necessary, and eliminates a substantial amount of effort currently required to provide feedback on any image. Assuming the fastest steps to accomplish this are currently 1- tap, 2- zoom (in my experience on smaller mobile screens at least), 3- tap yes/no… with or without a tap required to exit, providing feedback and returning to the feed of detections for review requires a minimum of three user actions.
  • This proposed enhancement will result in at least a 300% increase in human input to model training in cases where a user whose current effort remains the same but completes two additional tasks for every one task currently completed
  • The actual benefit would likely be far greater as a result of the dramatic reduction in effort combined with nudges to compel further engagement from users who receive instant gratification from completion of a task they found at least partially desireable to complete
  • Existing users will find greater satisfaction and are likely to invest greater effort towards greater fulfillment, yielding exponentially more training data
  • Potential users of this feature will increase, including those who do not engage as a matter of effort required, leading to an increase in overall user engagement
  • Less effort for all users decreases the risk to perceived opportunity cost and individual impediment to action, while providing far greater incentive in the form of increased and imminently available neurochemical reward we instinctively crave and produce as a species in general.
@rjfehr rjfehr added the enhancement New feature or request label Jan 26, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

1 participant