Pix2Text

Pix2Text copied to clipboard

找不到ConvBNActivation的声明是怎么回事？？

1

源代码： ``` from pix2text import Pix2Text img_fp = 'C:/Users/PC/AppData/Local/Programs/Python/Python38/Envs/Pytorch/formula.jpg' p2t = Pix2Text() out_text = p2t(img_fp) # 也可以使用 `p2t.recognize(img_fp)` 获得相同的结果 print(out_text) ``` 运行结果： cannot import name 'ConvBNActivation' from 'torchvision.models.mobilenetv2' 如图： ...

zongpingding

关于recognize函数

2

请问作者有没有考虑对recognize函数后续提供可以选择指定识别模型的参数呢，比如如果指定模型如type="formula"那么就选择指定公式模型进行识别，但是如果传入是"none"那么还是以前的自动识别。 因为如下图所示，对于同一个公式我发现如果截图的框选范围不同有时会导致在"**general**"和"**formula**"之间横跳  

Alune233

failed to download the models through cn oss

3

I have solved it mostly, though I have to make a warning for others' convenience: when first using pix2text and automatically downloading models, whether cnocr or cnstd used the url...

vandermondeDET

YoloV7_Tiny 自动下载失败

1

YoloV7_Tiny 自动下载失败 [WARNING 2024-04-07 17:15:53,577 _assert_and_prepare_model_files:214] Can NOT find model file C:\Users\Administrator\AppData\Roaming\cnstd\1.2\analysis\mfd-yolov7_tiny.pt [ERROR 2024-04-07 17:15:53,917 get_model_file:251] Failed downloading url https://sg-models.oss-cn-beijing.aliyuncs.com/cnstd/1.2/yolov7_tiny_mfd-pytorch.zip pix2text=1.0.2.3

xiehurricane

Index out of bounds when merging text

1

``` for _out in outs: line_number = _out.get('line_number', 0) + 1 if len(out_texts)

tonyzdev

可以添加文档识别功能

1

最近在逛github时看到https://github.com/Mathpix/mpx-cli这个项目，不知道本项目可不可以借鉴一下，做出类似的功能。

zsy1207

请问下分式 和大括号有时候检测不到，是参数不对吗？ 下面是hf-demo测试的结果

4

 

openforward

Patterns of incorrect response

1

Hey everyone, I was benchmarking Pix2Text with Mathpix and found the following patterns for which Pix2Text gives incorrect output while Mathpix gives the correct response. If there are plans for...

ishan00

docker部署到python:3.10.13会返回<unused3>

2

调用recognize(img, use_analyzer=True)返回json会带 { "code": 200, "message": "OK", "data": [ { "type": "embedding", "text": " $$\\frac{3}{5}$$ ", "position": [ [ 205.0, 11.0 ], [ 223.0, 11.0 ], [ 223.0, 49.0 ],...

FocusHuang

Idea : merge digital extraction result and OCR result.

5

Hi, I'm currently developing a pdf parser specialised for math pdf. The non-OCR solutions offer great accuracy for text because they are simply extracted, not detected optically. So, is it...

Shadow-Alex