ImageMonkey / imagemonkey-core

ImageMonkey is an attempt to create a free, public open source image dataset.
https://imagemonkey.io
47 stars 10 forks source link

ability to search for existing annotations in Unified (not quite the same as "rework") #279

Open dobkeratops opened 3 years ago

dobkeratops commented 3 years ago

I'd like to be able to add more detail to an image that has existing annotations (eg refining more parts on person or vehicle annotations, and adding kerbstones to street scenes)

.. there's currently no way to find such images in unified ; you can browse them in "explore", you can also find them in unified by selecting the "rework" option (which shows alot of candidates), but there you are not allowed to actually add new annotations :(

I think this would have to be an additional option or filter alongside "rework"?

Options: No options (o) Rework ( ) Highlight ( ) => Options: Pending tasks (o) Add new annotations to annotated image ( ) Rework existing annotations ( ) Highlight ( )

is there a better way to present it? (i think it's hard to explain the difference between rework and what i'm after)

Search for images // instead of "search for tasks you are interested in" - search for images.
Unlabelled [ ]  random [ ] 
specific labels: [  .... ]    
Unannotated[ ]   Annotated[ ]   
// possible to search for both. 
// .. default "un-annoted" has same effect as searching for pending tasks
// .. search for 'annotated' is similar to "rework", but see the idea for rework below

( The unified mode still seems to present itself as a task search.. it might be more elegant to just present it as an image search. it can do the job of "explore", or add labels, or annotate. In general you dont 'annotate tasks' .. you just browse images, then annotate what you see that needs doing immiediately)

Could rework be done better from within unified itself, i.e. whenever you see anything broken, you press a button to rework it?

eventually I think you can make the unified browser do the job of "expore" better - because once you're looking at the image, you can toggle through everything else there aswell.