Mobile interface

[dlangenk]

I thought we already talked about this but I couldn't find an issue about it.
What about a different view to the largo process. A sketch is attached.
Basically it shows you just one largo thumbnail and you can swipe or use the arrow keys to different directions to assign a label or discard the annotation.

• When you load the tool/view you can select a category and then the first thumbnail of that group is displayed in the middle.
• To the left the discard "label" is automatically selected, but this can be changed
• To the right the label selected in step 1 is selected, but this can be changed
• (Optionally) we can assign different labels to the top or bottom
• When you swipe / press the arrow key I could think of two things
  • Either the assignment is instant, i.e. the object is relabeled/discarded (or the current label is kept)
  • Or the selection is remembered and you can press continue and get to the current second step of Largo with the labels already attached to the thumbnails

[mzur]

Yup, I will work on something like this as part of my GEOMAR contract. My ideas are this:

MIRA (Mobile Interface to Relabel Annotations)

At first, users only annotate regularly using a single label (e.g. "interesting"). Then they can either use Largo on the Desktop or MIRA on a mobile device to relabel these annotations. This should be combined with LabelBOT which will be offered as a processing job similar to the laser point detection. Users select the initial label first ("Interesting") and LabelBOT will replace it with the most likely label it can find in the whole volume (let's ignore the issues with missing example annotations for now). MIRA will then be designed for an efficient review process of the LabelBOT labels. One gesture/swipe confirms the label. Another gesture opens the label list with the favorite/recently used labels to choose a different label for the annotation. A third gesture swaps the LabelBOT label with the label that was used last. The fourth gesture could be used to go back in case a mistake was made. These interactions will be instantly saved and the tool remembers the last position so users can easily resume. This should also make it possible to use the tool for a Largo-like workflow.