animl-frontend
animl-frontend copied to clipboard
Improve interactions with bounding boxes
Currently only clicking and dragging from left to right and top to bottom works. Make bottom to top and right to left work as well.
- [ ] Support clicking and dragging in all directions when drawing new bounding box
- [x] Make dragging to & around edges of image more bullet proof
- [ ] make grabbing small objects and moving them easier
- [ ] whole ux can get a little laggy (including cross hairs catching up to mouse when creating a bbox, the drawing of the bbox itself). Try to figure out a fix.
Also related, Roboflow has a "hand" tool for making these kinds of adjustments. I think the hand cursor is nice but don't think it needs to be it's own tool/mode, but might be worth taking another look at how they display corner handles and manage selecting objects etc.
They do a kind of nice thing in which, when in hand tool mode, a dark transparent overlay is applied to all background that's not a object rectangle, and an additional one gets applied when you hover over a specific object, making the other objects appear a little darker too and making it very clear which one you're selecting. I like that quite a bit.
- [ ] making validate/invalidate buttons easier to click to the labels.
- [ ] make it easier to select objects that sit below other objects. One idea for a fix: add "send to back / send to front" items to bounding-box context menus, the same way Google Slides and vector editing programs do it.
- [ ] single button to remove object (rather than having to invalidate whole stack of labels)?
- [ ] click-and-drag to select multiple objects?
We have also gotten feedback from a few users that it's not obvious that you can click the label on the bounding box to change (add) a new one, Folks were validating megadetctor objects then adding new objects around them for the species labels.
Also the unlock/lock button could use a hard look. One user was editing labels then unlocking the objects because he thought the unlock button was the lock button.
- [ ] button to hide all the bounding boxes to make it easier to see the image (Nat Geo user suggestion)
- [ ] make currently focused bounding box more clear
Received some feedback from users that the UX makes it seem like users are editing a current label, rather than adding an new one on top of it. Fair point.