-
Notifications
You must be signed in to change notification settings - Fork 40
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Support imports of common project formats #6
Comments
I'm unable to export to Darknet properly, all the .txt files are empty (0 bytes). Can you please provide an example (support I have two classes created in the software - cat & dog)? Also, thanks a ton, this is such a wonderful tool! |
Sure, I'll take a look. Can you try exporting to another format to see if you get any labels out? Bit odd that you're getting empty text files, that's normally the case when you have unlabelled images. So it's exporting the images OK? |
COCO works fine.
|
OK, just to check - are you importing or exporting? If import can you show me a label file too? (or even upload a zip of your project with some examples and I can try running it myself) |
Apologies, I assumed that setting the .names file for a detector makes use of it while exporting. There isn't a bug in the code. :) |
@harshalgajjar actually this should be the case, see here for example: deeplabel/src/darknetexporter.cpp Line 3 in ecc24c5
Setting the names file determines which labels get exported - it's important for the case when e.g. you label some images, and so does your friend, but you in your deeplabel project you didn't create the classes in the same order (say you have cat=1, dog=2 and they have dog=1, cat=2). That's only the case for darknet currently - though you've highlighted a good point. This should probably be the same for all the exporters that explicitly use class IDs; GCP for example uses the class name [string] so it's not an issue. |
Minimally should be able to import arbitrary projects in the following formats:
DarknetMS COCOThe text was updated successfully, but these errors were encountered: