add main.py

main.py

@@ -0,0 +1,46 @@
+
+from pix2text import Pix2Text, merge_line_texts
+
+img_fp = './examples/en1.jpg'
+p2t = Pix2Text.from_config()
+outs = p2t.recognize_text_formula(img_fp, resized_shape=768, return_text=True)
+print(outs)
+
+
+
+
+
+
+
+
+
+
+# #! pip install pix2text>=1.1
+# import requests
+# from io import BytesIO
+# from PIL import Image
+
+# from pix2text import Pix2Text, merge_line_texts
+
+# def download_img(url):
+#     response = requests.get(url)
+#     image_file = BytesIO(response.content)
+#     return Image.open(image_file).convert('RGB')
+
+# image_fps = [
+#     'https://raw.githubusercontent.com/breezedeus/Pix2Text/main/docs/examples/formula.jpg',
+#     'https://raw.githubusercontent.com/breezedeus/Pix2Text/main/docs/examples/math-formula-42.png',
+# ]
+# images = [download_img(img_fp) for img_fp in image_fps]
+
+# p2t = Pix2Text.from_config()
+# outs = p2t.recognize_formula(images, return_text=True)  # recognize pure formula images
+# print('outs for pure formula images', outs)
+
+# mixed_img_fp = 'https://raw.githubusercontent.com/breezedeus/Pix2Text/main/docs/examples/en1.jpg'
+# mixed_img = download_img(mixed_img_fp)
+
+# outs2 = p2t.recognize(mixed_img, file_type='text_formula', return_text=True, save_analysis_res='en1-out.jpg')  # recognize mixed images
+# print(outs2)
+# from IPython.display import display
+# display(Image.open('en1-out.jpg'))

readme.txt

@@ -1 +1,31 @@
-https://github.com/breezedeus/Pix2Text
+
+https://github.com/breezedeus/Pix2Text
+
+pip install pix2text
+    # base 环境
+
+p2t predict -l en,ch_sim --resized-shape 768 --file-type text_formula -i docs/examples/en1.jpg --save-debug-res out-debug-en1.jpg 
+    # 成功识别
+
+p2t predict -l en,ch_sim --mfd-config '{"model_name": "mfd-pro-1.5", "model_backend": "onnx"}' --formula-ocr-config '{"model_name":"mfr-pro-1.5","model_backend":"onnx"}' --text-ocr-config '{"rec_model_name": "doc-densenet_lite_666-gru_large"}' --resized-shape 768 --file-type text_formula -i docs/examples/en1.jpg --save-debug-res out-debug-en1.jpg
+    # 付费模型
+
+C:\Users\Administrator\AppData\Roaming\pix2text\1.1\table-rec\model.safetensors
+    # 模型保存在这了
+
+
+
+
+
+
+
+
+
+
+
+
+conda create formulaocr python==3.10 pip
+
+pip install -r .\requirements.txt
+
+

scripts/try_pix2text_mfr.py

@@ -63,7 +63,7 @@ def test_tokenizer_consistency(processor, test_strings=None):
     print("="*50 + "\n")
 
 model = 'breezedeus/pix2text-mfr'
-model = 'models/checkpoint-683356'
+
 processor = TrOCRProcessor.from_pretrained(model)
 
 # 测试Tokenizer的编码和解码是否一致
@@ -71,6 +71,7 @@ processor = TrOCRProcessor.from_pretrained(model)
 
 # model = ORTModelForVision2Seq.from_pretrained(model, use_cache=False)
 
+model = 'models/checkpoint-683356'
 model = VisionEncoderDecoderModel.from_pretrained(model)
 
 image_fps = [