Hugging Face's logo

Spaces:
pablocst
/
ocr_translate
Sleeping

App Files Community
ocr_translate
File size: 4,744 Bytes 
5a0f532
 
 
5eb66cf
 
5a0f532
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e16eaa7
 
 
 
 
 
 
 
 
 
 
5a0f532
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e16eaa7
 
 
 
 
 
 
 
 
 
 
5a0f532
e16eaa7
5a0f532
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
 
# OCR Translate v0.1
# 创建人：曾逸夫
# 创建时间：2022-06-14
# email: zyfiy1314@163.com
# 项目地址：https://gitee.com/CV_Lab/ocr-translate

import os

import gradio as gr
import nltk
import pytesseract
from nltk.tokenize import sent_tokenize
from transformers import MarianMTModel, MarianTokenizer

nltk.download('punkt')

OCR_TR_DESCRIPTION = '''# OCR Translate v0.1
<div id="content_align">基于Tesseract的OCR翻译系统</div>'''

# 图片路径
img_dir = "./data"

# 获取tesseract语言列表
choices = os.popen('tesseract --list-langs').read().split('\n')[1:-1]


# 翻译模型选择
def model_choice(src="en", trg="zh"):
    # https://huggingface.co/Helsinki-NLP/opus-mt-en-zh
    model_name = f"Helsinki-NLP/opus-mt-{src}-{trg}"  # 模型名称

    tokenizer = MarianTokenizer.from_pretrained(model_name)  # 分词器
    model = MarianMTModel.from_pretrained(model_name)  # 模型

    return tokenizer, model


# tesseract语言列表转pytesseract语言
def ocr_lang(lang_list):
    lang_str = ""
    lang_len = len(lang_list)
    if lang_len == 1:
        return lang_list[0]
    else:
        for i in range(lang_len):
            lang_list.insert(lang_len - i, "+")

        lang_str = "".join(lang_list[:-1])
        return lang_str


# ocr tesseract
def ocr_tesseract(img, languages):
    ocr_str = pytesseract.image_to_string(img, lang=ocr_lang(languages))
    return ocr_str


# 示例
def set_example_image(example: list) -> dict:
    return gr.Image.update(value=example[0])


# 清除
def clear_content():
    return None


# 翻译
def translate(input_text):
    # 参考：https://huggingface.co/docs/transformers/model_doc/marian
    if input_text is None or input_text == "":
        return "系统提示：没有可翻译的内容！"

    tokenizer, model = model_choice()

    translate_text = ""
    input_text_list = input_text.split("\n\n")

    for i in range(len(input_text_list)):
        translated_sub = model.generate(
            **tokenizer(sent_tokenize(input_text_list[i]), return_tensors="pt", truncation=True, padding=True))
        tgt_text_sub = [tokenizer.decode(t, skip_special_tokens=True) for t in translated_sub]
        translate_text_sub = "".join(tgt_text_sub)
        translate_text = translate_text + "\n\n" + translate_text_sub

    return translate_text[2:]


def main():

    with gr.Blocks(css='style.css') as ocr_tr:
        gr.Markdown(OCR_TR_DESCRIPTION)

        # -------------- OCR 文字提取 --------------
        with gr.Box():

            with gr.Row():
                gr.Markdown("### Step 01: 文字提取")

            with gr.Row():
                with gr.Column():
                    with gr.Row():
                        inputs_img = gr.Image(image_mode="RGB", source="upload", type="pil", label="图片")
                    with gr.Row():
                        inputs_lang = gr.CheckboxGroup(choices=choices, type="value", value=['eng'], label='语言')

                    with gr.Row():
                        clear_img_btn = gr.Button('Clear')
                        ocr_btn = gr.Button(value='OCR 提取', variant="primary")

                with gr.Column():
                    imgs_path = sorted(os.listdir(img_dir))
                    example_images = gr.Dataset(components=[inputs_img],
                                                samples=[[f"{img_dir}/{i}"] for i in imgs_path])

        # -------------- 翻译 --------------
        with gr.Box():

            with gr.Row():
                gr.Markdown("### Step 02: 翻译")

            with gr.Row():
                with gr.Column():
                    with gr.Row():
                        outputs_text = gr.Textbox(label="提取内容", lines=20)
                    with gr.Row():
                        clear_text_btn = gr.Button('Clear')
                        translate_btn = gr.Button(value='翻译', variant="primary")
                with gr.Column():
                    outputs_tr_text = gr.Textbox(label="翻译内容", lines=20)

        # ---------------------- OCR Tesseract ----------------------
        ocr_btn.click(fn=ocr_tesseract, inputs=[inputs_img, inputs_lang], outputs=[
            outputs_text,])

        clear_img_btn.click(fn=clear_content, inputs=[], outputs=[inputs_img])

        example_images.click(fn=set_example_image, inputs=[
            example_images,], outputs=[
                inputs_img,])

        # ---------------------- OCR Tesseract ----------------------
        translate_btn.click(fn=translate, inputs=[outputs_text], outputs=[outputs_tr_text])
        clear_text_btn.click(fn=clear_content, inputs=[], outputs=[outputs_text])

    ocr_tr.launch(inbrowser=True)


if __name__ == '__main__':
    main()