The text-recognition model was trained on approximately one million handwritten text-line images ranging from the 17th to the 19th century. See the model card for detailed evaluation results, and results from some fine-tuning experiments.
