Update app.py

app.py

@@ -1,52 +1,40 @@
-import re
-import jaconv
-import gradio as gr
-from transformers import AutoTokenizer, AutoFeatureExtractor, VisionEncoderDecoderModel
-from PIL import Image
-import torch
-import cv2
 import os
 os.system('pip install paddlepaddle')
 os.system('pip install paddleocr')
 from paddleocr import PaddleOCR, draw_ocr
+from PIL import Image
+import gradio as gr
+import torch
 
-tokenizer = AutoTokenizer.from_pretrained("kha-white/manga-ocr-base")
-
-model = VisionEncoderDecoderModel.from_pretrained("kha-white/manga-ocr-base")
-
-feature_extractor = AutoFeatureExtractor.from_pretrained("kha-white/manga-ocr-base")
-
-examples = ["japan.jpg"]
-
-def post_process(text):
-  text = ''.join(text.split())
-  text = text.replace('…', '...')
-  text = re.sub('[・.]{2,}', lambda x: (x.end() - x.start()) * '.', text)
-  text = jaconv.h2z(text, ascii=True, digit=True)
-  return text
-
-def manga_ocr(img):
-  ocr = PaddleOCR(use_angle_cls=True, lang='japan',use_gpu=False)
-  img_path = img.name
-  result = ocr.ocr(img_path, cls=True)
-  image = Image.open(img_path).convert('RGB')
-  pixel_values = feature_extractor(img, return_tensors="pt").pixel_values
-  output = model.generate(pixel_values)[0]
-  text = tokenizer.decode(output, skip_special_tokens=True)
-  text = post_process(text)
-  return text
-
-iface = gr.Interface(
-    fn=manga_ocr,
-    inputs=[gr.inputs.Image(label="Input", type="pil")],
-    outputs="text",
-    layout="horizontal",
-    theme="huggingface",
-    title="Manga OCR",
-    description="Optical Character Recognization for Japanese Texts with focus on Mangas. The model is trained by kha-white with Github link: <a href=\"https://github.com/kha-white/manga-ocr\">manga-ocr</a> while the Space App is made by me.",
-    allow_flagging='never',
+torch.hub.download_url_to_file('https://i.imgur.com/aqMBT0i.jpg', 'example.jpg')
+
+def inference(img, lang):
+    ocr = PaddleOCR(use_angle_cls=True, lang=lang,use_gpu=False)
+    img_path = img.name
+    result = ocr.ocr(img_path, cls=True)
+    image = Image.open(img_path).convert('RGB')
+    boxes = [line[0] for line in result]
+    txts = [line[1][0] for line in result]
+    scores = [line[1][1] for line in result]
+    im_show = draw_ocr(image, boxes, txts, scores,
+                       font_path='simfang.ttf')
+    im_show = Image.fromarray(im_show)
+    im_show.save('result.jpg')
+    return 'result.jpg'
+
+title = 'PaddleOCR'
+description = 'Gradio demo for PaddleOCR. PaddleOCR demo supports Chinese, English, French, German, Korean and Japanese.To use it, simply upload your image and choose a language from the dropdown menu, or click one of the examples to load them. Read more at the links below.'
+article = "<p style='text-align: center'><a href='https://www.paddlepaddle.org.cn/hub/scene/ocr'>Awesome multilingual OCR toolkits based on PaddlePaddle （practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices）</a> | <a href='https://github.com/PaddlePaddle/PaddleOCR'>Github Repo</a></p>"
+examples = [['example.jpg','en']]
+css = ".output_image, .input_image {height: 40rem !important; width: 100% !important;}"
+gr.Interface(
+    inference,
+    [gr.inputs.Image(type='file', label='Input'),gr.inputs.Dropdown(choices=['ch', 'en', 'fr', 'german', 'korean', 'japan'], type="value", default='en', label='language')],
+    gr.outputs.Image(type='file', label='Output'),
+    title=title,
+    description=description,
+    article=article,
     examples=examples,
-    article = "Author: <a href=\"https://huggingface.co/gryan-galario\">Gryan Galario</a>",
-)
-
-iface.launch()
+    css=css,
+    enable_queue=True
+    ).launch(debug=True)