## 背景 经过需求征集https://github.com/PaddlePaddle/PaddleOCR/issues/10334 和每周技术研讨会 https://github.com/PaddlePaddle/PaddleOCR/issues/10223 讨论,我们确定了【论文复现】OCR识别模型Parseq任务。 ## 解决步骤 1. 根据开源代码进行网络结构、评估指标转换。代码链接:https://github.com/baudm/parseq 2. 结合[论文复现指南](https://github.com/PaddlePaddle/models/blob/release%2F2.2/tutorials/article-implementation/ArticleReproduction_CV.md)和[新增模型指南](https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/doc/doc_ch/add_new_algorithm.md),进行前反向对齐等操作,达到论文Table.6中的指标。 3. 参考[PR提交规范](https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/doc/doc_ch/code_and_doc.md)提交代码PR到ppocr中。