Starstruk Logo MarkStarstruk

v2.6.0

PaddlePaddle/PaddleOCRv2.6.0Aug 24, 2022by MissPenguin

AI Summary

This release introduces PP-Structurev2 with significant upgrades for Chinese scenes, including Layout Recovery and a PDF-to-Word conversion feature. It also includes major optimizations for Layout Analysis, Table Recognition, and Key Information Extraction, resulting in substantial improvements in model storage, speed, and accuracy.

Key Highlights

  • Release of PP-Structurev2 with Layout Recovery and one-line PDF to Word conversion.
  • Layout Analysis optimization: 95% reduction in model storage and 11x speed increase.
  • Table Recognition optimization: 6% accuracy improvement using 3 new strategies.
  • KIE optimization: 9.1% improvement in relation extraction accuracy.

New Features

  • PDF to Word conversion
  • Layout Recovery
  • Visual-independent KIE model structure
  • Optimized Table Recognition model

Full Release Notes

**Release Note**
- Release [PP-Structurev2](./ppstructure/)，with functions and performance fully upgraded, adapted to Chinese scenes, and new support for [Layout Recovery](./ppstructure/recovery) and **one line command to convert PDF to Word**;
- [Layout Analysis](./ppstructure/layout) optimization: model storage reduced by 95%, while speed increased by 11 times, and the average CPU time-cost is only 41ms;
- [Table Recognition](./ppstructure/table) optimization: 3 optimization strategies are designed, and the model accuracy is improved by 6% under comparable time consumption;
- [Key Information Extraction](./ppstructure/kie) optimization：a visual-independent model structure is designed, the accuracy of semantic entity recognition is increased by 2.8%, and the accuracy of relation extraction is increased by 9.1%.