dive icon indicating copy to clipboard operation
dive copied to clipboard

Advanced Species Selection Tool

Open kelseylmartin opened this issue 4 weeks ago • 0 comments

Is your feature request related to a problem? If so, Please describe. Currently, our readers are running through the videos as "viame-assisted" readers where they are using the annotations to assist in their manual counts of the species in the video. At the beginning of each read, a reader will skim through the video and check off any species they see in the video to trim down annotated species list. They will then skip to the "MaxN Frame" for each species and double check that the annotations are correct within that frame and move on the next species. This becomes a problem when the model has incorrectly guessed a species that our readers have not checked and therefore a track does not show up for a fish that appears on the screen. They then either have to create a new track or go back and click through the species that were not selected, both of which can be quite time-consuming.

Describe the solution you'd like Add a "drag and drop" box for selecting species. Above each box is a button that you can toggle on and off for species that you have selected vs species that you have not. Will thoroughly help our QA/QC process and prevent readers from having to go through entire species list if model incorrectly identifies a species.

Describe alternatives you've considered Instead of a drag and drop box that could potentially clutter the UI even more with already limited space, it could also just be a separate little check box next to each species itself but with the same "on/off" feature above the whole species list.

Additional context Would there also be a way to add a "human confirmed" flag to the annotations that have been manually confirmed by a reader? Should I create a separate issue for this? I was thinking you select a track or a group of tracks and then check a box that says "human confirmed" and then this gets flagged as a separate code in the track csv (i.e., not 1.0 but "hc" or some unique designation so I can filter for it in the data analysis)

kelseylmartin avatar Dec 01 '25 21:12 kelseylmartin