(brainstorming - summarised annotation mode suggestions - mockup)

[dobkeratops]

ideas shown:

• larger work area:-

  • trim the 'instruction text' (would be ok to say "annotate all: car" on the same line.. but I was sort of hoping the 'top menu' already tells you?)

  • move the toolbar to the left side (widescreen leverage)

• Precise rectangle outlines - 3 alternative ideas shown:-

  • (i)transparent area fill with opaque single-pixel outline,

  • (ii)similar but with only the corners highlighted (bin not car in the image)

  • (iii) or thick outline (more like the current) but transparent, and the precise outline drawn single pixel outline. The intent is to clearly show the precise single-pixel boundary

• 'plural' flag: default could be 'unspecified' i.e. you may annotate groups or individuals (a later quiz can confirm). The ideal is of course a specific toggle (car/cars)

• label drop-box (allowing switch to another label.. such as head or hands of person in the close-up images)

[dobkeratops]

(regarding the Smart Annotation mode - did you consider overlaying the smart-annotation boundary over the same image? having said that regardless, I can think of many other uses for a 'side-by-side' mode, so it's great to have)

[bbernhard]

Many thanks for the suggestions - really like those!

At the moment I am working on the "larger work area" part. As this already is quite a big change (had to restructure some code in order to clean up the code base a bit - I really hate Javascript ;)), I think I am going to incrementally push those changes to production.

One thing I am unsure at the moment is: the image size. On my working git branch, I tried to make the image as big as possible. While this works great for large monitors, it's not that nice of a experience on a small laptop (-> image doesn't fit anymore on a page, so you have to scroll).

Ideally the image size should be calculated depending on the free space, but I guess this is quite hard with the framework I am currently using. So I am thinking about introducing three different modes (small/middle/large workspace), where "small" would match the current size. It's something that could be selected in the "annotation settings".

(regarding the Smart Annotation mode - did you consider overlaying the smart-annotation boundary over the same image? having said that regardless, I can think of many other uses for a 'side-by-side' mode, so it's great to have)

yeah, I wanted it to do it that way originally. But it turned out, that it gets pretty confusing when there are a lot of scribbles - at some point you can't distuingish between scribbles and the polygon that gets returned from the smart annotation.

[dobkeratops]

. While this works great for large monitors, it's not that nice of a experience on a small laptop (-> image doesn't fit anymore on a page, so you have to scroll).

yeah I can see that. i'm often using a 13" laptop. I guess the best is if you dont need scrolling when the window is maximised

So I am thinking about introducing three different modes (small/middle/large workspace), where "small" would match the current size. It's something that could be selected in the "annotation settings".

makes sense. You might want to use a smaller window to one side whilst doing something else, who knows..

But it turned out, that it gets pretty confusing when there are a lot of scribbles

fair enough.