Repository navigation
A user should only be presented with results that are accurate and useful #890
Description
Summary
The ultimate predictions that are displayed in the various views should be accurate enough for a users needs. This means using post-processing techniques, filtering bad data and presenting warnings in the UI so that it is clear what data is reliable and what is not.
Implementation Details
This includes things such as:
- Filtering crops that are under a certain size [Draft] Filter images before clustering #858
- Rolling up low confidence predictions to a higher rank [Draft] Taxon rank classifications ("rank roll-ups") #857
- Exclude "Not a Moth" by default #868
- Make score threshold a project setting #873
- Complete support for detection tracking Add support for occurrence tracking #863
- Class masking (world model to regional moths) Support for "geo-fencing" approach to improve prediction accuracy for regional species list #797
Activity
[-]A user should be presented with results that are accurate and useful[/-][+]A user should only be presented with results that are accurate and useful[/+]