Launching today
Dataset-Maker
Tear documents on purpose, with the answer key attached.
1 follower
Tear documents on purpose, with the answer key attached.
1 follower
Dataset-Maker turns PDFs into non-overlapping torn fragments with exact ground truth for placement and adjacency. It also ships dm-bench, a reproducible torn-document reassembly benchmark with easy, medium and hard tiers, an evaluation harness, and a deterministic classical baseline. Built for document-reassembly research, with reproducibility and verifiable ground truth at the core.
