Using the Viewer, I spend a lot of time moving and resizing manual annotations to tag everyone of interest in each image, even if their face is too small to be automatically detected or if it is turned away from the camera. Currently annotations are placed with the same size and location and always needs to be moved and resized to the desired face.
This request is for enhancements to the placement of manual face annotations to make it more efficient. Rather than placing them automatically, IMatch would wait for user mouse input to define position and (optionally) size. Behavior could be:
- user clicks a location -> annotation placed at default size centered on click
- user clicks and drags across a face and releases -> annotation placed with location and size bounded by click and release points
- user Ctrl-clicks from center of image to desired corner location -> annotation placed with location and size bounded by click and release points
+1
+1
+1 - like the click/drag approach for defining regions/areas.
+1
+1
Sounds like something I should look into for IMatch 2021.
+1 from my side also: I would love to have that click&drag approach (also in combination with SHIFT to keep the face annotation in square ratio)
At the moment I'm doing some face-tagging session and usually the default face annotation size is much too large. And copying/moving/resizing/renaming existing face annotations (what I use now) is also not the fastest and sometimes introduces some inconsistencies within the DB (fixed by DB Diagnostics).
Under which conditions (steps) does this create inconsistencies. I also work that way and have never seen problems. I also use the annotation gallery to store face annotations for people (saves a lot of copy/paste).
Difficult to tell when it exactly happens = reproduce it. I see regularly warnings like this in DB Diagnostics:
Warning: Face region found in file[241534] with invalid face link [47297]. Face link removed.
Warning: File [242981] face [46173] has a mismatch in face/anno sort order.
What I e.g. also do is to copy 2-3 face annotations at once (e.g. when I have a squence of photos of my family) and then move them around.
Also it could happen that I manualy delete one of the annotations again when I have double-face annotations for same person (e.g. face recognition found a face in the series)
Plus a small thing. One can create face annotation over the buttom and right border of the picture (but not over the other two). Why?
One must not create face annotation outside the picture.
+1