search

Particle1904 / DatasetHelpers

Dataset Helper program to automatically select, re scale and tag Datasets (composed of image and text) for Machine Learning training.
MIT License
170 stars 9 forks source link

[Feature Request]: Improved keyboard navigation #24

Closed Cruxial0 closed 5 months ago

Cruxial0 commented 5 months ago

Been using this program for editing my datasets recently. It's great, however I feel it lacks in terms of keyboard navigation. With tag editing being a very keyboard intensive task, it makes sense to allow the user to perform some actions using the keyboard.

Suggested keyboard shortcuts (Tag/Caption Editor) (Windows/Linux):

EDIT: ALT + Arrow Keys is already implemented, oops. As a quality of life improvement to this bind, the caret should be set to either the end of the text, or in a position relative to where you were on the previous page.

EDIT 2: Sometimes when using the ALT + Arrow Keys keybind, the letter 'B' is added to the start of the prompt. It seemingly only happens after editing text THEN swapping pages.

Particle1904 commented 5 months ago

EDIT 2: Sometimes when using the ALT + Arrow Keys keybind, the letter 'B' is added to the start of the prompt. It seemingly only happens after editing text THEN swapping pages.

~I noticed that and I think its a bug in the keyboard library I'm using since it didn't happen when I first implemented the feature (but I updated the dependency to a new version).~ Doesn't look like its happening anymore in the current commit.

Particle1904 commented 5 months ago

I've changed it so the caret position stays in the same position after changing images (in terms of characters count from the beginning of the text) or snaps to the end if the caret position is bigger than the last image text size. It will also focus on the Editor control to avoid users having to manually click on it to continue editing text. These features will be available in the next release v2.8.0.

~Seems like the random "B" being added is still occurring in the current dev commit and I have no idea why. Its extremely hard to debug that since its so random. I think its something to do with Alt + Codes... I've noticed it also adding an "X", a "," and even a "▬"~. I tried:

It seems like this is a default feature of Avalonia, I found only a single issue in their GitHub about this and there's a single response from another user. Basically if you type any text into a text Control then do any type of Alt Code combination, it adds the "special" character, for example typing "a" then pressing Alt + 2 + 5 + 5 adds a "7" in the text box... as a workaround, F1 to F6 keys (and Mouse with Key 4 and Key 5) can be used to navigate between images as described here.

I have no plans to support editing of weighted captions/tags, so Alt + Up and Alt + Down won't be added. I added generating tags with weights a while back and never tried training a model with it, and since its something that's not that popular/common, I wont add support.