Add Belege (synthetic German invoices) to Important Datasets - #29
Open
immineal wants to merge 1 commit into
Open
Add Belege (synthetic German invoices) to Important Datasets#29immineal wants to merge 1 commit into
immineal wants to merge 1 commit into
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds one dataset to the Important Datasets section: Belege, a synthetic set of 1,000 German invoices/credit notes with pixel-level bounding-box labels for every field and line item.
Why it might fit here: the list currently has no vision/document dataset, only text/LLM-oriented ones. Belege covers three structurally different page layouts and the two German zero-tax scenarios (§19 UStG small-business exemption, §13b UStG reverse charge) that a model trained on English invoices or SROIE typically mishandles.
Disclosure: I made this dataset. A free 40-document sample is on Hugging Face/Kaggle/Zenodo (linked in the entry); the full 1,000-document set is a paid download, also linked. Happy to adjust wording, trim the entry, or close this if it does not fit the list's scope - it's the first non-text-generation dataset proposed here, so a reasonable call either way.