Is your feature request related to a problem? Please describe.
In cases where you have not been collecting the data yourself, you might come across a situation where you start labeling data and later discover that the initial assumption of number of individuals is wrong.
You then need to update the config.yml with the additional individuals, and then force an update of the already labeled frames. This is doable if it happens after a few videos, but I just had this happen after 55 videos, and the process is very tedious.
As outlined in #2125 there was (and in my experience still is) the issue that when you update the config.yml, the label frame GUI (when initiated via the main GUI) is not pulling the labels from the config, but from the h5 files already associated with the frames. I can confirm that this issue is still existing in rc14.
Describe the solution you'd like
Ideally there would be a function like dlc.multianimal_add_additional(config, [individual10, individual11, individual12]) that would add additional individuals to the config and then force an update on all previously labeled frames.
As a bonus, it should reset the training iteration to 0 and prevent attempts to merge, because trying to merge training data with different number of individuals causes errors as I have learned.
If that is beyond the scope, it would be great to at least point that out in the documentation as a potential error source.
Describe alternatives you've considered
Similar to the solution suggested here #2723 (comment), I loaded napari, activated the dlc plugin, load the folder with the frames, then load the config, click "no" in the dialog, save the layer, remove the layer and frames, and move to the next folder.
I wrote a companion script to run a pre-check of of my labeled frames against the config to guide me what still needs updating and to confirm that everything has been saved correctly, but it still left the manual loading, saving, etc.
I tried my hand at adding some kind of automation, i.e. loading napari with the correct config attached, but I just ended up with crashing napari repeatedly
Additional context
The manual approach works fine (even though it is not very straightforward) for a few videos, but with more videos it becomes tedious and time consuming.
Is your feature request related to a problem? Please describe.
In cases where you have not been collecting the data yourself, you might come across a situation where you start labeling data and later discover that the initial assumption of number of individuals is wrong.
You then need to update the
config.ymlwith the additional individuals, and then force an update of the already labeled frames. This is doable if it happens after a few videos, but I just had this happen after 55 videos, and the process is very tedious.As outlined in #2125 there was (and in my experience still is) the issue that when you update the
config.yml, the label frame GUI (when initiated via the main GUI) is not pulling the labels from the config, but from the h5 files already associated with the frames. I can confirm that this issue is still existing in rc14.Describe the solution you'd like
Ideally there would be a function like
dlc.multianimal_add_additional(config, [individual10, individual11, individual12])that would add additional individuals to the config and then force an update on all previously labeled frames.As a bonus, it should reset the training iteration to 0 and prevent attempts to merge, because trying to merge training data with different number of individuals causes errors as I have learned.
If that is beyond the scope, it would be great to at least point that out in the documentation as a potential error source.
Describe alternatives you've considered
Similar to the solution suggested here #2723 (comment), I loaded napari, activated the dlc plugin, load the folder with the frames, then load the config, click "no" in the dialog, save the layer, remove the layer and frames, and move to the next folder.
I wrote a companion script to run a pre-check of of my labeled frames against the config to guide me what still needs updating and to confirm that everything has been saved correctly, but it still left the manual loading, saving, etc.
I tried my hand at adding some kind of automation, i.e. loading napari with the correct config attached, but I just ended up with crashing napari repeatedly
Additional context
The manual approach works fine (even though it is not very straightforward) for a few videos, but with more videos it becomes tedious and time consuming.