Hello Ines,
Couple of questions here:
1- If we want to modify the line opacity or the line width of a bounding box so it becomes thinner, is this something we can do now with the current version by adding something to our config
recipe file or prodigy.json
file? I tried adding the bounding_box_opacity
and other options found on this thread in my prodigy.json
but it didn't work out.
2- Is there a way to also remove the polygon and the freehand tool from appearing on the UI.
3- Will there be like an argument to pass by to the command line of db-out
to remove the base64
data from the exported database to the jsonl file? Maybe something like --remove-b64
.
Thank you for your constant support.
G
You can find the up-to-date customization options for the UI here: Annotation interfaces · Prodigy · An annotation tool for AI, Machine Learning & NLP
You can customize the line width by setting image_manual_stroke_width
. If you want to change the opacity, you can just use a semi-transparent color as the hex color for the given label, e.g. #00000030
. This also lets you set different opacity values for different labels if needed. You can overwrite the label colors via the custom_theme
config setting.
You can configure the available annotation modes by setting the image_manual_modes
config setting.
I'd say this is a bit too specific for the general-purpose db-out
– the data could be anything in theory, the base64 encoded value could live under any key, and there could be other types of data that a user may want to remove or edit as well. In the meantime, it should be pretty simple to do the same in only a few lines of code. For example:
from prodigy.components.db import connect
import srsly
db = connect()
examples = db.get_dataset("your_dataset")
for eg in examples:
eg["image"] = ""
srsly.write_jsonl("exported_data.jsonl", examples)
(If you know jq
, there's probably a simple one-liner you can use on the CLI to do the same thing. But I couldn't tell you how to do it )