Will Paddle release a lightweight recognition model that uses ppocrv4_doc_dict.txt as dictionary?

[brianyuastri]

Hi everyone,

ch_PP-OCRv4_server_rec_doc was released recently (~3 weeks ago)
https://github.com/PaddlePaddle/PaddleOCR/blob/main/docs/ppocr/model_list.md

and it works wonderfully on a significantly larger range of characters (15000+)
https://github.com/PaddlePaddle/PaddleOCR/blob/33fa33e4a990bdef095a4a74cf849781312c2b67/ppocr/utils/dict/ppocrv4_doc_dict.txt

I wonder if there are any plans from Paddle to release a lightweight version of the recognition model that uses the same set of character dictionary. Thank you!

[GreatV]

As of now, there is no direct mention in the provided documents regarding the release of a lightweight recognition model using ppocrv4_doc_dict.txt as its character dictionary. However, PaddleOCR has been actively developing and releasing new recognition models, including multilingual and ultra-lightweight models.

Given that ch_PP-OCRv4_server_rec_doc was recently released and supports an extensive character set (15,000+), it is possible that a lightweight version could be considered in future updates. It would be best to follow the official discussion forums and model list updates for any announcements. You can check the PaddleOCR model list and related discussions here:

➡️ PaddleOCR Model List
➡️ PaddleOCR Discussions

If you require a lightweight model with an extended character set, you might consider fine-tuning an existing lightweight model using custom training with ppocrv4_doc_dict.txt.

---

Response generated by 🤖 feifei-bot | chatgpt-4o-latest

⚠️ This is an AI-generated response. Please verify any critical information.