openocr-python 0.1.0.dev0__tar.gz → 0.1.0.dev2__tar.gz

{openocr_python-0.1.0.dev0/openocr_python.egg-info → openocr_python-0.1.0.dev2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: openocr-python
-Version: 0.1.0.dev0
+Version: 0.1.0.dev2
 Summary: Accurate and Efficient General OCR System
 Home-page: https://github.com/Topdu/OpenOCR
 Author: OCR Team, FVL Lab
@@ -27,35 +27,29 @@ Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: numpy==1.26.4
-Requires-Dist: opencv-python<=4.12.0
-Requires-Dist: Pillow>=8.0.0
-Requires-Dist: pyyaml>=5.4.0
-Requires-Dist: tqdm>=4.60.0
-Requires-Dist: rapidfuzz>=2.0.0
-Requires-Dist: pyclipper>=1.3.0
+Requires-Dist: numpy
+Requires-Dist: opencv-python
+Requires-Dist: Pillow
+Requires-Dist: pyyaml
+Requires-Dist: tqdm
+Requires-Dist: rapidfuzz
+Requires-Dist: pyclipper
 Requires-Dist: pydantic
 Requires-Dist: shapely
-Provides-Extra: onnx
-Requires-Dist: onnxruntime>=1.12.0; extra == "onnx"
+Requires-Dist: huggingface-hub
+Requires-Dist: modelscope
+Requires-Dist: onnxruntime
 Provides-Extra: onnx-gpu
-Requires-Dist: onnxruntime-gpu>=1.12.0; extra == "onnx-gpu"
+Requires-Dist: onnxruntime-gpu; extra == "onnx-gpu"
 Provides-Extra: pytorch
-Requires-Dist: torch>=1.10.0; extra == "pytorch"
-Requires-Dist: torchvision>=0.11.0; extra == "pytorch"
+Requires-Dist: torch; extra == "pytorch"
+Requires-Dist: torchvision; extra == "pytorch"
 Provides-Extra: gradio
-Requires-Dist: gradio==4.20.0; extra == "gradio"
-Provides-Extra: huggingface
-Requires-Dist: huggingface-hub>=0.16.0; extra == "huggingface"
-Provides-Extra: modelscope
-Requires-Dist: modelscope>=1.9.0; extra == "modelscope"
+Requires-Dist: gradio; extra == "gradio"
 Provides-Extra: all
-Requires-Dist: onnxruntime>=1.12.0; extra == "all"
-Requires-Dist: torch>=1.10.0; extra == "all"
-Requires-Dist: torchvision>=0.11.0; extra == "all"
-Requires-Dist: gradio==4.20.0; extra == "all"
-Requires-Dist: huggingface-hub>=0.16.0; extra == "all"
-Requires-Dist: modelscope>=1.9.0; extra == "all"
+Requires-Dist: torch; extra == "all"
+Requires-Dist: torchvision; extra == "all"
+Requires-Dist: gradio; extra == "all"
 Dynamic: author
 Dynamic: home-page
 Dynamic: license-file
@@ -81,13 +75,13 @@ Visit: https://github.com/Topdu/OpenOCR
 
 ```bash
 # Install from PyPI (recommended)
-pip install openocr-python huggingface-hub==0.36.0 modelscope onnxruntime
+pip install openocr-python==0.1.0.dev2
 
 # Or install from source
 git clone https://github.com/Topdu/OpenOCR.git
 cd OpenOCR
 python build_package.py
-pip install ./build/dist/openocr_python-*.whl huggingface-hub==0.36.0 modelscope onnxruntime
+pip install ./build/dist/openocr_python-*.whl
 ```
 
 ## Command Line Usage
@@ -173,7 +167,7 @@ openocr --task doc --input_path path/to/img --use_layout_detection --save_vis --
 
 ```bash
 # Install gradio
-pip install gradio==4.20.0
+pip install gradio
 ```
 
 ### OCR Demo
@@ -181,7 +175,6 @@ pip install gradio==4.20.0
 Launch Gradio web interface for OCR tasks:
 
 ```bash
-pip install onnxruntime
 # Local access only
 openocr --task launch_openocr_demo --server_port 7860
 
@@ -194,7 +187,6 @@ openocr --task launch_openocr_demo --server_port 7860 --share
 Launch Gradio web interface for universal recognition:
 
 ```bash
-pip install onnxruntime
 openocr --task launch_unirec_demo --server_port 7861 --share
 ```
 
@@ -203,7 +195,6 @@ openocr --task launch_unirec_demo --server_port 7861 --share
 Launch Gradio web interface for document parsing:
 
 ```bash
-pip install onnxruntime
 openocr --task launch_opendoc_demo --server_port 7862 --share
 ```
 

{openocr_python-0.1.0.dev0 → openocr_python-0.1.0.dev2}/QUICKSTART.md

@@ -18,13 +18,13 @@ Visit: https://github.com/Topdu/OpenOCR
 
 ```bash
 # Install from PyPI (recommended)
-pip install openocr-python huggingface-hub==0.36.0 modelscope onnxruntime
+pip install openocr-python==0.1.0.dev2
 
 # Or install from source
 git clone https://github.com/Topdu/OpenOCR.git
 cd OpenOCR
 python build_package.py
-pip install ./build/dist/openocr_python-*.whl huggingface-hub==0.36.0 modelscope onnxruntime
+pip install ./build/dist/openocr_python-*.whl
 ```
 
 ## Command Line Usage
@@ -110,7 +110,7 @@ openocr --task doc --input_path path/to/img --use_layout_detection --save_vis --
 
 ```bash
 # Install gradio
-pip install gradio==4.20.0
+pip install gradio
 ```
 
 ### OCR Demo
@@ -118,7 +118,6 @@ pip install gradio==4.20.0
 Launch Gradio web interface for OCR tasks:
 
 ```bash
-pip install onnxruntime
 # Local access only
 openocr --task launch_openocr_demo --server_port 7860
 
@@ -131,7 +130,6 @@ openocr --task launch_openocr_demo --server_port 7860 --share
 Launch Gradio web interface for universal recognition:
 
 ```bash
-pip install onnxruntime
 openocr --task launch_unirec_demo --server_port 7861 --share
 ```
 
@@ -140,7 +138,6 @@ openocr --task launch_unirec_demo --server_port 7861 --share
 Launch Gradio web interface for document parsing:
 
 ```bash
-pip install onnxruntime
 openocr --task launch_opendoc_demo --server_port 7862 --share
 ```
 

{openocr_python-0.1.0.dev0 → openocr_python-0.1.0.dev2}/openocr/__init__.py

@@ -13,7 +13,7 @@ sys.path.insert(0, os.path.abspath(os.path.join(__dir__, '..')))
 # from .tools.infer_doc_onnx import OpenDocONNX
 from .openocr import OpenOCR, main
 
-__version__ = '0.1.0.dev'
+__version__ = '0.1.0.dev2'
 
 # Lazy import for demo interfaces to avoid initialization on import
 def launch_openocr_demo(*args, **kwargs):

{openocr_python-0.1.0.dev0 → openocr_python-0.1.0.dev2}/openocr/demo_gradio.py

@@ -1,7 +1,7 @@
 # @Author: OpenOCR
 # @Contact: 784990967@qq.com
 import os
-import gradio as gr  # gradio==4.20.0
+import gradio as gr
 
 os.environ['FLAGS_allocator_strategy'] = 'auto_growth'
 import cv2
@@ -9,6 +9,8 @@ import numpy as np
 import json
 import time
 from PIL import Image
+from pathlib import Path
+
 from tools.infer_e2e import OpenOCRE2E, check_and_download_font, draw_ocr_box_txt
 from tools.download_example_images import get_example_images_path
 
@@ -218,8 +220,8 @@ def launch_demo(share=False, server_port=7860, server_name='0.0.0.0'):
                                  img_output,
                                  img_mask,
                              ])
-
-    demo.launch(share=share, server_port=server_port, server_name=server_name)
+    allowed_path = str(Path.home() / '.cache' / 'openocr')
+    demo.launch(share=share, server_port=server_port, server_name=server_name, allowed_paths=[allowed_path])
     return demo
 
 

{openocr_python-0.1.0.dev0 → openocr_python-0.1.0.dev2}/openocr/demo_opendoc.py

@@ -6,7 +6,10 @@ import base64
 import argparse
 import gradio as gr
 from PIL import Image
+from pathlib import Path
 
+import sys
+sys.path.append(os.path.dirname(os.path.abspath(__file__)))
 from tools.infer_doc_onnx import OpenDocONNX
 from tools.utils.logging import get_logger
 from tools.download_example_images import get_example_images_path
@@ -496,10 +499,13 @@ def launch_demo(
 
     demo = create_demo()
     # Launch with settings from arguments
+    allowed_path = str(Path.home() / '.cache' / 'openocr')
+    # Launch demo
     demo.launch(
         share=share,
+        server_name=server_name,
         server_port=server_port,
-        server_name=server_name
+        allowed_paths=[allowed_path]
     )
     return demo
 

{openocr_python-0.1.0.dev0 → openocr_python-0.1.0.dev2}/openocr/demo_unirec.py

@@ -2,9 +2,7 @@ import re
 import gradio as gr
 import numpy as np
 from PIL import Image
-from threading import Thread
-import queue
-import time
+from pathlib import Path
 
 # Import ONNX inference components
 import sys
@@ -75,15 +73,30 @@ model = None
 
 
 # --- 2. Streaming generation function ---
-def stream_generate(inference, image, max_length=2048, result_queue=None):
-    """Generate text with streaming output."""
+def stream_recognize_image(input_image, max_length=2048):
+    """Stream recognition results with dual display: markdown text during recognition, render after completion.
+
+    Args:
+        input_image: Input PIL Image
+        max_length: Maximum generation length
+    """
+    if input_image is None:
+        yield '请先上传一张图片。', '**请先上传一张图片。**'
+        return
+
+    # Convert to PIL Image if needed
+    if not isinstance(input_image, Image.Image):
+        input_image = Image.fromarray(input_image).convert('RGB')
+    else:
+        input_image = input_image.convert('RGB')
+
     # Get token IDs
-    bos_token_id = inference.tokenizer.bos_token_id
-    eos_token_id = inference.tokenizer.eos_token_id
-    pad_token_id = inference.tokenizer.pad_token_id
+    bos_token_id = model.tokenizer.bos_token_id
+    eos_token_id = model.tokenizer.eos_token_id
+    pad_token_id = model.tokenizer.pad_token_id
 
     # Encode image
-    encoder_hidden_states, cross_k, cross_v = inference.encode_image(image)
+    encoder_hidden_states, cross_k, cross_v = model.encode_image(input_image)
 
     # Initialize generation
     generated_ids = [bos_token_id]
@@ -91,23 +104,24 @@ def stream_generate(inference, image, max_length=2048, result_queue=None):
     # Initialize empty past_key_values
     batch_size = encoder_hidden_states.shape[0]
     past_key_values = []
-    for _ in range(inference.num_decoder_layers):
+    for _ in range(model.num_decoder_layers):
         empty_key = np.zeros(
-            (batch_size, inference.num_heads, 0, inference.head_dim),
+            (batch_size, model.num_heads, 0, model.head_dim),
             dtype=np.float32)
         empty_value = np.zeros(
-            (batch_size, inference.num_heads, 0, inference.head_dim),
+            (batch_size, model.num_heads, 0, model.head_dim),
             dtype=np.float32)
         past_key_values.append((empty_key, empty_value))
+
     cleaned_text = ''
-    put_token_num = 30
+
     # Generation loop with streaming
     for step in range(max_length - 1):
         current_token = generated_ids[-1]
         past_length = step
 
         # Decode step
-        logits, past_key_values = inference.decode_step(
+        logits, past_key_values = model.decode_step(
             current_token,
             past_length,
             cross_k,
@@ -120,81 +134,22 @@ def stream_generate(inference, image, max_length=2048, result_queue=None):
         next_token_id = int(np.argmax(logits[0, -1, :]))
         generated_ids.append(next_token_id)
 
-        # Decode current sequence and put in queue
-        if result_queue is not None:
-            current_text = inference.tokenizer.decode(generated_ids[-1:], skip_special_tokens=False)
-            # print(current_text+'\n')
-            cleaned_text = cleaned_text + clean_special_tokens(current_text)
-            # Post-process HTML table attributes
-            # cleaned_text = cleaned_text.replace('<tdcolspan=', '<td colspan=')
-            # cleaned_text = cleaned_text.replace('<tdrowspan=', '<td rowspan=')
-            # cleaned_text = cleaned_text.replace('"colspan=', '" colspan=')
-            if (step + 1) % put_token_num == 0:
-                result_queue.put(cleaned_text)
-        result_queue.put(cleaned_text)
+        # Decode and clean current token
+        current_text = model.tokenizer.decode(generated_ids[-1:], skip_special_tokens=False)
+        cleaned_text += clean_special_tokens(current_text)
+
+        yield cleaned_text, '_正在识别中，请稍候..._'
+
         # Check for EOS
         if next_token_id == eos_token_id:
             break
 
-    # Signal completion
-    if result_queue is not None:
-        result_queue.put(None)
-
-
-# --- 3. Gradio streaming function for dual display ---
-def stream_recognize_image(input_image):
-    """Stream recognition results with dual display: markdown text only during recognition, render after completion."""
-    if input_image is None:
-        yield '请先上传一张图片。', '**请先上传一张图片。**'
-        return
-
-    # Convert to PIL Image if needed
-    if not isinstance(input_image, Image.Image):
-        input_image = Image.fromarray(input_image).convert('RGB')
+    # Final yield - render the complete result
+    if cleaned_text:
+        formatted_result = format_markdown_output(cleaned_text)
+        yield formatted_result, formatted_result
     else:
-        input_image = input_image.convert('RGB')
-
-    # Create queue for streaming results
-    result_queue = queue.Queue()
-
-    # Start generation in background thread
-    thread = Thread(target=stream_generate, args=(model, input_image, 2048, result_queue))
-    thread.daemon = True  # Set as daemon thread
-    thread.start()
-
-    # Stream results - only update markdown text, keep render area with "recognizing" message
-    last_update_time = time.time()
-    current_text = ''
-
-    while True:
-        try:
-            # Get result with longer timeout
-            result = result_queue.get(timeout=1.0)
-            if result is None:  # Generation complete
-                break
-            current_text = result
-            last_update_time = time.time()
-
-            # Only update markdown text, show "recognizing" message in render area
-            yield current_text, '_正在识别中，请稍候..._'
-
-        except queue.Empty:
-            # No new result yet, check if thread is still alive
-            if not thread.is_alive():
-                # Thread finished but no completion signal, break
-                break
-            # Yield current state periodically to keep UI responsive
-            current_time = time.time()
-            if current_time - last_update_time > 0.5:  # Update UI every 0.5s
-                yield current_text if current_text else '正在识别中...', '_正在识别中，请稍候..._'
-                last_update_time = current_time
-
-    # Wait for thread to complete
-    thread.join(timeout=2.0)
-
-    # Final yield - now render the complete result
-    formatted_result = format_markdown_output(current_text)
-    yield formatted_result, formatted_result
+        yield '识别失败，请重试。', '**识别失败，请重试。**'
 
 
 def format_markdown_output(markdown_text):
@@ -223,69 +178,71 @@ def format_markdown_output(markdown_text):
     # Gradio's Markdown component will handle the rendering
     return markdown_text
 
-
-# --- 4. Gradio Interface ---
-# Get example images path and download if necessary
-example_img_dir = get_example_images_path(demo_type='unirec')
-
-# Get list of example images
-example_images = []
-if os.path.exists(example_img_dir):
-    for file in os.listdir(example_img_dir):
-        if file.lower().endswith(('.png', '.jpg', '.jpeg', '.bmp', '.gif')):
-            example_images.append(os.path.join(example_img_dir, file))
-    example_images = sorted(example_images)
-
-with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.HTML("""
-            <h1 style='text-align: center;'><a href="https://github.com/Topdu/OpenOCR">UniRec-0.1B: Unified Text and Formula Recognition with 0.1B Parameters</a></h1>
-            <p style='text-align: center;'>0.1B超轻量模型统一文本与公式识别（由<a href="https://fvl.fudan.edu.cn">FVL实验室</a> <a href="https://github.com/Topdu/OpenOCR">OCR Team</a> 创建）</p>
-            <p style='text-align: center;'><a href="https://github.com/Topdu/OpenOCR/blob/main/docs/unirec.md">[本地GPU部署]</a>获取快速识别体验</p>"""
-            )
-    gr.Markdown('上传一张图片，点击"运行识别"按钮进行文本和公式识别。')
-    with gr.Row():
-        with gr.Column(scale=4):  # 左侧竖排：图片 + 按钮
-            image_input = gr.Image(label='上传图片 or 粘贴截图', type='pil')
-
-            # Add examples if available
-            if example_images:
-                gr.Examples(
-                    examples=example_images,
-                    inputs=image_input,
-                    label='📚 示例图片'
+def create_demo() -> gr.Blocks:
+
+    # --- 4. Gradio Interface ---
+    # Get example images path and download if necessary
+    example_img_dir = get_example_images_path(demo_type='unirec')
+
+    # Get list of example images
+    example_images = []
+    if os.path.exists(example_img_dir):
+        for file in os.listdir(example_img_dir):
+            if file.lower().endswith(('.png', '.jpg', '.jpeg', '.bmp', '.gif')):
+                example_images.append(os.path.join(example_img_dir, file))
+        example_images = sorted(example_images)
+
+    with gr.Blocks(theme=gr.themes.Soft()) as demo:
+        gr.HTML("""
+                <h1 style='text-align: center;'><a href="https://github.com/Topdu/OpenOCR">UniRec-0.1B: Unified Text and Formula Recognition with 0.1B Parameters</a></h1>
+                <p style='text-align: center;'>0.1B超轻量模型统一文本与公式识别（由<a href="https://fvl.fudan.edu.cn">FVL实验室</a> <a href="https://github.com/Topdu/OpenOCR">OCR Team</a> 创建）</p>
+                <p style='text-align: center;'><a href="https://github.com/Topdu/OpenOCR/blob/main/docs/unirec.md">[本地GPU部署]</a>获取快速识别体验</p>"""
                 )
+        gr.Markdown('上传一张图片，点击"运行识别"按钮进行文本和公式识别。')
+        with gr.Row():
+            with gr.Column(scale=4):  # 左侧竖排：图片 + 按钮
+                image_input = gr.Image(label='上传图片 or 粘贴截图', type='pil')
+
+                # Add examples if available
+                if example_images:
+                    gr.Examples(
+                        examples=example_images,
+                        inputs=image_input,
+                        label='📚 示例图片'
+                    )
+
+                with gr.Row():
+                    run_button = gr.Button('🚀 运行识别', variant='primary')
+                    clear_button = gr.Button('🗑️ 清空', variant='secondary')
+
+            with gr.Column(scale=6):
+                with gr.Tabs():
+                    with gr.Tab('📄 Markdown Source'):
+                        markdown_output = gr.Code(label='Markdown Source',
+                                                language='markdown',
+                                                lines=20)
+                    with gr.Tab('📝 Markdown Preview'):
+                        markdown_render = gr.Markdown(
+                                value='_渲染后的表格/公式将显示在这里..._',
+                                latex_delimiters=LATEX_DELIMS,
+                                elem_id='md_preview')
+
+        # 点击运行按钮后触发
+        run_button.click(
+            stream_recognize_image,
+            inputs=[image_input],
+            outputs=[markdown_output, markdown_render]
+        )
 
-            with gr.Row():
-                run_button = gr.Button('🚀 运行识别', variant='primary')
-                clear_button = gr.Button('🗑️ 清空', variant='secondary')
-
-        with gr.Column(scale=6):
-            with gr.Tabs():
-                with gr.Tab('📝 Markdown Source'):
-                    markdown_output = gr.Code(label='Markdown Source',
-                                            language='markdown',
-                                            lines=20)
-                with gr.Tab('📝 Markdown Preview'):
-                    markdown_render = gr.Markdown(
-                            value='_渲染后的表格/公式将显示在这里..._',
-                            latex_delimiters=LATEX_DELIMS,
-                            elem_id='md_preview')
-
-    # 点击运行按钮后触发
-    run_button.click(
-        stream_recognize_image,
-        inputs=[image_input],
-        outputs=[markdown_output, markdown_render]
-    )
-
-    # 清空按钮功能：清空图片和输出结果
-    def clear_all():
-        return None, '', '_渲染后的表格/公式将显示在这里..._'
+        # 清空按钮功能：清空图片和输出结果
+        def clear_all():
+            return None, '', '_渲染后的表格/公式将显示在这里..._'
 
-    clear_button.click(
-        clear_all,
-        outputs=[image_input, markdown_output, markdown_render]
-    )
+        clear_button.click(
+            clear_all,
+            outputs=[image_input, markdown_output, markdown_render]
+        )
+    return demo
 
 
 def launch_demo(
@@ -323,12 +280,15 @@ def launch_demo(
         use_gpu=use_gpu,
         auto_download=auto_download
     )
+    demo = create_demo()
 
+    allowed_paths = str(Path.home() / '.cache' / 'openocr')
     # Launch demo
-    demo.queue().launch(
+    demo.launch(
         share=share,
         server_name=server_name,
-        server_port=server_port
+        server_port=server_port,
+        allowed_paths=[allowed_paths]
     )
     return demo
 

{openocr_python-0.1.0.dev0 → openocr_python-0.1.0.dev2}/openocr/tools/infer_doc_onnx.py

@@ -999,7 +999,7 @@ class OpenDocONNX:
         elif 'table' in base_label:
             f.write(f'{merged_text}\n\n')
         elif 'formula' in base_label or base_label == 'equation':
-            f.write(f'$${merged_text}$$\n\n')
+            f.write(f'{merged_text}\n\n')
         else:
             f.write(f'{merged_text}\n\n')
 

{openocr_python-0.1.0.dev0 → openocr_python-0.1.0.dev2/openocr_python.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: openocr-python
-Version: 0.1.0.dev0
+Version: 0.1.0.dev2
 Summary: Accurate and Efficient General OCR System
 Home-page: https://github.com/Topdu/OpenOCR
 Author: OCR Team, FVL Lab
@@ -27,35 +27,29 @@ Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: numpy==1.26.4
-Requires-Dist: opencv-python<=4.12.0
-Requires-Dist: Pillow>=8.0.0
-Requires-Dist: pyyaml>=5.4.0
-Requires-Dist: tqdm>=4.60.0
-Requires-Dist: rapidfuzz>=2.0.0
-Requires-Dist: pyclipper>=1.3.0
+Requires-Dist: numpy
+Requires-Dist: opencv-python
+Requires-Dist: Pillow
+Requires-Dist: pyyaml
+Requires-Dist: tqdm
+Requires-Dist: rapidfuzz
+Requires-Dist: pyclipper
 Requires-Dist: pydantic
 Requires-Dist: shapely
-Provides-Extra: onnx
-Requires-Dist: onnxruntime>=1.12.0; extra == "onnx"
+Requires-Dist: huggingface-hub
+Requires-Dist: modelscope
+Requires-Dist: onnxruntime
 Provides-Extra: onnx-gpu
-Requires-Dist: onnxruntime-gpu>=1.12.0; extra == "onnx-gpu"
+Requires-Dist: onnxruntime-gpu; extra == "onnx-gpu"
 Provides-Extra: pytorch
-Requires-Dist: torch>=1.10.0; extra == "pytorch"
-Requires-Dist: torchvision>=0.11.0; extra == "pytorch"
+Requires-Dist: torch; extra == "pytorch"
+Requires-Dist: torchvision; extra == "pytorch"
 Provides-Extra: gradio
-Requires-Dist: gradio==4.20.0; extra == "gradio"
-Provides-Extra: huggingface
-Requires-Dist: huggingface-hub>=0.16.0; extra == "huggingface"
-Provides-Extra: modelscope
-Requires-Dist: modelscope>=1.9.0; extra == "modelscope"
+Requires-Dist: gradio; extra == "gradio"
 Provides-Extra: all
-Requires-Dist: onnxruntime>=1.12.0; extra == "all"
-Requires-Dist: torch>=1.10.0; extra == "all"
-Requires-Dist: torchvision>=0.11.0; extra == "all"
-Requires-Dist: gradio==4.20.0; extra == "all"
-Requires-Dist: huggingface-hub>=0.16.0; extra == "all"
-Requires-Dist: modelscope>=1.9.0; extra == "all"
+Requires-Dist: torch; extra == "all"
+Requires-Dist: torchvision; extra == "all"
+Requires-Dist: gradio; extra == "all"
 Dynamic: author
 Dynamic: home-page
 Dynamic: license-file
@@ -81,13 +75,13 @@ Visit: https://github.com/Topdu/OpenOCR
 
 ```bash
 # Install from PyPI (recommended)
-pip install openocr-python huggingface-hub==0.36.0 modelscope onnxruntime
+pip install openocr-python==0.1.0.dev2
 
 # Or install from source
 git clone https://github.com/Topdu/OpenOCR.git
 cd OpenOCR
 python build_package.py
-pip install ./build/dist/openocr_python-*.whl huggingface-hub==0.36.0 modelscope onnxruntime
+pip install ./build/dist/openocr_python-*.whl
 ```
 
 ## Command Line Usage
@@ -173,7 +167,7 @@ openocr --task doc --input_path path/to/img --use_layout_detection --save_vis --
 
 ```bash
 # Install gradio
-pip install gradio==4.20.0
+pip install gradio
 ```
 
 ### OCR Demo
@@ -181,7 +175,6 @@ pip install gradio==4.20.0
 Launch Gradio web interface for OCR tasks:
 
 ```bash
-pip install onnxruntime
 # Local access only
 openocr --task launch_openocr_demo --server_port 7860
 
@@ -194,7 +187,6 @@ openocr --task launch_openocr_demo --server_port 7860 --share
 Launch Gradio web interface for universal recognition:
 
 ```bash
-pip install onnxruntime
 openocr --task launch_unirec_demo --server_port 7861 --share
 ```
 
@@ -203,7 +195,6 @@ openocr --task launch_unirec_demo --server_port 7861 --share
 Launch Gradio web interface for document parsing:
 
 ```bash
-pip install onnxruntime
 openocr --task launch_opendoc_demo --server_port 7862 --share
 ```