-
I got a mail asking if mask refinement is also working. I think this would greatly improve the usability of SAM. If we are touching the UI already other functionality could be added as well (don't kno…
-
Dear Martin,
whilst scanning a large image with the lawnmower mode to make sure we do not miss anything, we often need to zoom in for investigating details of individual objects. Alas, lawnmower mo…
-
Investigate how Mask R-CNN classification performs and if it can be optionally enabled for MAIA. This would require additional steps during training proposal selection (see #42). However, if existing …
-
If the video annotation timeline contains some thousand annotation elements, it becomes rather slow to interact with (especially zooming and panning). Consider to implement the timeline as one single …
-
Maybe we could replace the manual as primary resource for help with a new "Help" portal (which replaces "Manual" in the top menu). The portal shows links to the community discussion forum (here on Git…
mzur updated
11 months ago
-
### General
- [ ] Prepare scaling plots until end of february. Y-axis: the speedup we get when running one epoch through the model for 2,4,6,8,10 GPUs
- [x] Find out how many samples we have in the …
-
Big images and small objects don't mix well. SAM is using 1024x1024 pixels as input size. If you have a 6000x8000px image small objects are hard to detect. A way around this is computing an embedding …
-
https://github.com/biigle/core/pull/893 changed the x-axis of the timeline chart from years to months. This makes it more likely that there are gaps between months where no annotations were created. T…
-
Some images might have a rotation flag set and it is neglected in some libraries but not in all, which leads to different orientations of the images used. E.g. in https://biigle.de/images/2240959/anno…
-
We had a request if BIIGLE could support more file metadata formats. These formats would require more than a simple mapping between column names as it is currently done. Some formats may also be very …