Search papers, labs, and topics across Lattice.
1
0
3
This study examines how synthetic and real historical training data should be combined for low-resource OCR, and finds that joint and sequential training yield broadly similar archival accuracy under the tested practical pipelines.
