Modelpublic

BDRC/GoogleBooks_C_v1

sourceHugging Facecc-by-4.0updated 2y agoView on Hugging Face

0likes13downloads

Model Card

An OCR model trained on the GoogleBooks Dataset using a CRNN architecture. We use the following architecture identifiers:

—E: Easter2 (https://huggingface.co/BDRC/GoogleBooksEv1)
—C: CRNN (this model)
—T: TrOCR -> version 1 (https://huggingface.co/BDRC/GoogleBooksTv1) and version 2 (https://huggingface.co/BDRC/GoogleBooksTv2). Version two is the recommended checkpoint if you want to run inferences. It was trained a little longer using the same data distribution. We keep v1 in oder to evaluate the actual differences on the predictions.