Wang Xin
Wang Xin
@madmalkav, That makes sense.
I don't think that PR being merged will work properly with the previous model.
Maybe we could try to keep the version information of the character dictionary (a new column in the model list is used to show the dictionary used by the model),...
@rycont PP-OCRV5 already supports Japanese recognition. You can refer to [text_recognition](https://www.paddleocr.ai/latest/en/version3.x/module_usage/text_recognition.html), and the dictionary file can be found at [ppocr/utils/dict/ppocrv5_dict.txt](https://github.com/PaddlePaddle/PaddleOCR/blob/release/3.2/ppocr/utils/dict/ppocrv5_dict.txt.)
https://github.com/PaddlePaddle/PaddleOCR/blob/main/doc/doc_ch/kie.md#21-%E5%90%AF%E5%8A%A8%E8%AE%AD%E7%BB%83 按照文档可以跑通了，用提供的数据集。
【报名】：3-4
Thank you for your suggestion! That sounds great, we'll support this feature in future versions.
The main issue is that in LabelMe, there is no direct association between rectangles (bounding boxes) and points (keypoints). This lack of linkage means you cannot easily determine which keypoints...
What are your runtime environment details, such as paddle and paddleocr versions?
Hi @omeruth, could you provide an example image so I can replicate the result?