app.py

import gradio as gr
import pandas as pd
import numpy as np
import easyocr
import torch
import cv2
import PIL
import sys 
import os
import uuid
from PIL import ImageFont, ImageDraw, Image
this=gr.Interface.load("spaces/Omnibus/Translate-100")


lang_id = {
    "":"",
    "Afrikaans": "af",
    "Albanian": "sq",
    "Amharic": "am",
    "Arabic": "ar",
    "Armenian": "hy",
    "Asturian": "ast",
    "Azerbaijani": "az",
    "Bashkir": "ba",
    "Belarusian": "be",
    "Bulgarian": "bg",
    "Bengali": "bn",
    "Breton": "br",
    "Bosnian": "bs",
    "Burmese": "my",
    "Catalan": "ca",
    "Cebuano": "ceb",
    "Chinese": "zh",
    "Chinese (simplified)": "zh",
    "Chinese (traditional)": "zh",    
    "Croatian": "hr",
    "Czech": "cs",
    "Danish": "da",
    "Dutch": "nl",
    "English": "en",
    "Estonian": "et",
    "Fulah": "ff",
    "Finnish": "fi",
    "French": "fr",
    "Western Frisian": "fy",
    "Gaelic": "gd",
    "Galician": "gl",
    "Georgian": "ka",
    "German": "de",
    "Greek": "el",
    "Gujarati": "gu",
    "Hausa": "ha",
    "Hebrew": "he",
    "Hindi": "hi",
    "Haitian": "ht",
    "Hungarian": "hu",
    "Irish": "ga",
    "Indonesian": "id",
    "Igbo": "ig",
    "Iloko": "ilo",
    "Icelandic": "is",
    "Italian": "it",
    "Japanese": "ja",
    "Javanese": "jv",
    "Kazakh": "kk",
    "Central Khmer": "km",
    "Kannada": "kn",
    "Korean": "ko",
    "Luxembourgish": "lb",
    "Ganda": "lg",
    "Lingala": "ln",
    "Lao": "lo",
    "Lithuanian": "lt",
    "Latvian": "lv",
    "Malagasy": "mg",
    "Macedonian": "mk",
    "Malayalam": "ml",
    "Mongolian": "mn",
    "Marathi": "mr",
    "Malay": "ms",
    "Nepali": "ne",
    "Norwegian": "no",
    "Northern Sotho": "ns",
    "Occitan": "oc",
    "Oriya": "or",
    "Panjabi": "pa",
    "Persian": "fa",
    "Polish": "pl",
    "Pushto": "ps",
    "Portuguese": "pt",
    "Romanian": "ro",
    "Russian": "ru",
    "Sindhi": "sd",
    "Sinhala": "si",
    "Slovak": "sk",
    "Slovenian": "sl",
    "Spanish": "es",
    "Somali": "so",
    "Serbian": "sr",
    "Serbian (cyrillic)": "sr",
    "Serbian (latin)": "sr",    
    "Swati": "ss",
    "Sundanese": "su",
    "Swedish": "sv",
    "Swahili": "sw",
    "Tamil": "ta",
    "Thai": "th",
    "Tagalog": "tl",
    "Tswana": "tn",
    "Turkish": "tr",
    "Ukrainian": "uk",
    "Urdu": "ur",
    "Uzbek": "uz",
    "Vietnamese": "vi",
    "Welsh": "cy",
    "Wolof": "wo",
    "Xhosa": "xh",
    "Yiddish": "yi",
    "Yoruba": "yo",
    "Zulu": "zu",
}
ocr_id = {
    "":"",
    "Afrikaans": "af",
    "Albanian": "sq",
    "Arabic": "ar",
    "Azerbaijani": "az",
    "Belarusian": "be",
    "Bulgarian": "bg",
    "Bengali": "bn",
    "Bosnian": "bs",
    "Chinese (simplified)": "ch_sim",
    "Chinese (traditional)": "ch_tra",
    "Croatian": "hr",
    "Czech": "cs",
    "Danish": "da",
    "Dutch": "nl",
    "English": "en",
    "Estonian": "et",
    "French": "fr",
    "German": "de",
    "Irish": "ga",
    "Hindi": "hi",
    "Hungarian": "hu",
    "Indonesian": "id",
    "Icelandic": "is",
    "Italian": "it",
    "Japanese": "ja",
    "Kannada": "kn",
    "Korean": "ko",
    "Lithuanian": "lt",
    "Latvian": "lv",
    "Mongolian": "mn",
    "Marathi": "mr",
    "Malay": "ms",
    "Nepali": "ne",
    "Norwegian": "no",
    "Occitan": "oc",
    "Polish": "pl",
    "Portuguese": "pt",
    "Romanian": "ro",
    "Russian": "ru",
    "Serbian (cyrillic)": "rs_cyrillic",
    "Serbian (latin)": "rs_latin",
    "Slovak": "sk",
    "Slovenian": "sl",
    "Spanish": "es",
    "Swedish": "sv",
    "Swahili": "sw",
    "Tamil": "ta",
    "Thai": "th",
    "Tagalog": "tl",
    "Turkish": "tr",
    "Ukrainian": "uk",
    "Urdu": "ur",
    "Uzbek": "uz",
    "Vietnamese": "vi",
    "Welsh": "cy",
    "Zulu": "zu",
}


def blur_im(img,bounds,target_lang,trans_lang,ocr_sens,font_fac):
    im = cv2.imread(img)
    im = cv2.cvtColor(im, cv2.COLOR_BGR2RGB)
   
    for bound in bounds:
        if bound[2]>=(ocr_sens):
            p0, p1, p2, p3 = bound[0]
            x = int(p0[0])
            y = int(p0[1])
            w = int(p2[0]) - int(x)
            h = int(p2[1]) - int(y)
            kernel = np.ones((3, 3), np.uint8)
            im[y:y+h, x:x+w] = cv2.dilate(im[y:y+h, x:x+w], kernel, iterations=3)
            im[y:y+h, x:x+w] = cv2.GaussianBlur(im[y:y+h, x:x+w],(51,51),0)
        else:
            pass
    im = Image.fromarray(im)
    for bound in bounds:
        if bound[2]>=(ocr_sens):
            p0, p1, p2, p3 = bound[0]
            x = int(p0[0])
            y = int(p0[1])
            w = int(p2[0]) - int(x)
            h = int(p2[1]) - int(y)
            draw = ImageDraw.Draw(im)
            text = this(bound[1],target_lang,trans_lang)
            font_size=int(int(h)*font_fac)
            font = ImageFont.truetype("./fonts/unifont-15.0.01.ttf", int(font_size))
            draw.text((x, y),text, font = font, fill=(0,0,0))
        else:
            pass
    return im
    
def draw_boxes(image, bounds, ocr_sens,width=1):
    draw = ImageDraw.Draw(image)
    for bound in bounds:
        if bound[2]>=(ocr_sens):
            color = "blue"
        else:
            color = "red"
        p0, p1, p2, p3 = bound[0]
        draw.line([*p0, *p1, *p2, *p3, *p0], fill=color, width=width)
    return image

def detect(img, target_lang,trans_lang,ocr_sens,font_fac,target_lang2=None):
    if target_lang2 != None and target_lang2 != "":
        lang=f"{lang_id[target_lang]}"
        lang2=f"{lang_id[target_lang2]}"
        lang=[lang,lang2]
    else:
        lang=[f"{ocr_id[target_lang]}"]
        pass

    img = Image.open(img)
    img.thumbnail((1000,1000), Image.Resampling.LANCZOS)
    path = f"/tmp/{uuid.uuid4()}.jpg"
    img.save(path)
    img1 = np.array(img)
    reader = easyocr.Reader(lang)
    bounds = reader.readtext(img1)
    im = PIL.Image.open(path)
    im_out=draw_boxes(im, bounds,ocr_sens)
    blr_out=blur_im(path,bounds,target_lang,trans_lang,ocr_sens,font_fac)
    return im_out,blr_out,pd.DataFrame(bounds),pd.DataFrame(bounds).iloc[:,1:]
    
with gr.Blocks() as robot:
    gr.Markdown("""<h1><center>Translate Image to Image</center></h1><h4><center>EasyOCR and facebook/m2m100_1.2B</center></h4><h7><center>*translation may not be accurate</center></h7>""")
    with gr.Accordion(label="Description",open=False):
        with gr.Row():
            gr.Markdown("""<p> Drop your image in the "Image to Translate" Box<br>
                           Select the Language to Detect in the Image<br>
                           Select the Language to Translate to<br>
                           Click "Go" Button </p>""")
            gr.Markdown("""<p> Translation Model: <a href="https://huggingface.co/facebook/m2m100_1.2B">facebook/m2m100_1.2B</a><br>
                               OCR: <a href="https://www.jaided.ai/easyocr">easyocr</a><br>
                            Influence by:<br> 
                            <a href="https://huggingface.co/spaces/venz/AW-05-GR-NLP-Image2Text-Multilingual-OCR">venz/AW-05-GR-NLP-Image2Text-Multilingual-OCR</a><br>
                            <a href="https://huggingface.co/spaces/Iker/Translate-100-languages">Iker/Translate-100-languages</a><br>
                            </p>""")


    with gr.Row():
        with gr.Column():
            im=gr.Image(label="Image to Translate",type="filepath")

        with gr.Column():
            with gr.Group():
                with gr.Row():
                    with gr.Column():
                        target_lang = gr.Dropdown(label="Detect language:", choices=list(ocr_id.keys()),value="English")
                        trans_lang = gr.Dropdown(label="Translate to:", choices=list(lang_id.keys()),value="Chinese")
                    with gr.Column():
                        ocr_sens=gr.Slider(0.1, 1, step=0.05,value=0.25,label="Detect Min Confidence")
                        font_fac=gr.Slider(0.1, 1, step =0.1,value=0.4,label="Font Scale")
                        ocr_space=gr.Slider(1,10, step=1,value=5,label="Future Function")            
            go_btn=gr.Button("Go")
    with gr.Row():
        with gr.Column():
            out_im=gr.Image()
        with gr.Column():
            trans_im=gr.Image()            

    with gr.Row():
        out_txt=gr.Textbox(lines=8)
        data_f=gr.Dataframe()
                        
    go_btn.click(detect,[im,target_lang,trans_lang,ocr_sens,font_fac],[out_im,trans_im,out_txt,data_f])
    #go_btn.click(detect,[im,target_lang,target_lang2],[out_im,trans_im,out_txt,data_f])
robot.queue(concurrency_count=10).launch()