Update README.md
Browse files
README.md
CHANGED
|
@@ -1,10 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
-
base_model: baidu/ERNIE-4.5-0.3B-Paddle
|
| 3 |
-
language:
|
| 4 |
-
- en
|
| 5 |
-
- zh
|
| 6 |
-
- multilingual
|
| 7 |
-
library_name: PaddleOCR
|
| 8 |
license: apache-2.0
|
| 9 |
pipeline_tag: image-text-to-text
|
| 10 |
tags:
|
|
@@ -17,8 +78,13 @@ tags:
|
|
| 17 |
- table
|
| 18 |
- formula
|
| 19 |
- chart
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
---
|
| 21 |
-
|
| 22 |
<div align="center">
|
| 23 |
|
| 24 |
|
|
@@ -28,15 +94,11 @@ PaddleOCR-VL: Boosting Multilingual Document Parsing via a 0.9B Ultra-Compact Vi
|
|
| 28 |
|
| 29 |
</h1>
|
| 30 |
|
| 31 |
-
**📖 Paper**: [PaddleOCR-VL: Boosting Multilingual Document Parsing via a 0.9B Ultra-Compact Vision-Language Model](https://huggingface.co/papers/2510.14528)
|
| 32 |
-
**💻 Code**: [PaddlePaddle/PaddleOCR GitHub](https://github.com/PaddlePaddle/PaddleOCR)
|
| 33 |
-
**🌐 Project Page**: [PaddleOCR-VL Documentation](https://www.paddleocr.ai/latest/en/version3.x/algorithm/PaddleOCR-VL/PaddleOCR-VL.html)
|
| 34 |
-
|
| 35 |
[](https://github.com/PaddlePaddle/PaddleOCR)
|
| 36 |
-
[](https://discord.gg/JPmZXDsEEK)
|
| 41 |
[](https://x.com/PaddlePaddle)
|
| 42 |
[](./LICENSE)
|
|
@@ -56,13 +118,13 @@ PaddleOCR-VL: Boosting Multilingual Document Parsing via a 0.9B Ultra-Compact Vi
|
|
| 56 |
|
| 57 |
### **Core Features**
|
| 58 |
|
| 59 |
-
1.
|
| 60 |
|
| 61 |
|
| 62 |
-
2.
|
| 63 |
|
| 64 |
|
| 65 |
-
3.
|
| 66 |
|
| 67 |
|
| 68 |
### **Model Architecture**
|
|
@@ -105,7 +167,6 @@ Python API usage:
|
|
| 105 |
|
| 106 |
```python
|
| 107 |
from paddleocr import PaddleOCRVL
|
| 108 |
-
|
| 109 |
pipeline = PaddleOCRVL()
|
| 110 |
output = pipeline.predict("https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/paddleocr_vl_demo.png")
|
| 111 |
for res in output:
|
|
@@ -125,7 +186,6 @@ for res in output:
|
|
| 125 |
--network host \
|
| 126 |
ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddlex-genai-vllm-server
|
| 127 |
```
|
| 128 |
-
|
| 129 |
2. Call the PaddleOCR CLI or Python API:
|
| 130 |
|
| 131 |
```bash
|
|
@@ -134,10 +194,8 @@ for res in output:
|
|
| 134 |
--vl_rec_backend vllm-server \
|
| 135 |
--vl_rec_server_url http://127.0.0.1:8080/v1
|
| 136 |
```
|
| 137 |
-
|
| 138 |
```python
|
| 139 |
from paddleocr import PaddleOCRVL
|
| 140 |
-
|
| 141 |
pipeline = PaddleOCRVL(vl_rec_backend="vllm-server", vl_rec_server_url="http://127.0.0.1:8080/v1")
|
| 142 |
output = pipeline.predict("https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/paddleocr_vl_demo.png")
|
| 143 |
for res in output:
|
|
@@ -147,7 +205,6 @@ for res in output:
|
|
| 147 |
```
|
| 148 |
|
| 149 |
**For more usage details and parameter explanations, see the [documentation](https://www.paddleocr.ai/latest/en/version3.x/pipeline_usage/PaddleOCR-VL.html).**
|
| 150 |
-
|
| 151 |
## Performance
|
| 152 |
|
| 153 |
### Page-Level Document Parsing
|
|
@@ -300,4 +357,4 @@ If you find PaddleOCR-VL helpful, feel free to give us a star and citation.
|
|
| 300 |
primaryClass={cs.CV},
|
| 301 |
url={https://arxiv.org/abs/2510.14528},
|
| 302 |
}
|
| 303 |
-
```
|
|
|
|
| 1 |
+
Hugging Face's logo
|
| 2 |
+
Hugging Face
|
| 3 |
+
Models
|
| 4 |
+
Datasets
|
| 5 |
+
Spaces
|
| 6 |
+
Community
|
| 7 |
+
Docs
|
| 8 |
+
Pricing
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
PaddlePaddle
|
| 12 |
+
/
|
| 13 |
+
PaddleOCR-VL
|
| 14 |
+
|
| 15 |
+
like
|
| 16 |
+
504
|
| 17 |
+
|
| 18 |
+
Following
|
| 19 |
+
PaddlePaddle
|
| 20 |
+
430
|
| 21 |
+
Image-Text-to-Text
|
| 22 |
+
PaddleOCR
|
| 23 |
+
Safetensors
|
| 24 |
+
English
|
| 25 |
+
Chinese
|
| 26 |
+
multilingual
|
| 27 |
+
ERNIE4.5
|
| 28 |
+
PaddlePaddle
|
| 29 |
+
image-to-text
|
| 30 |
+
ocr
|
| 31 |
+
document-parse
|
| 32 |
+
layout
|
| 33 |
+
table
|
| 34 |
+
formula
|
| 35 |
+
chart
|
| 36 |
+
|
| 37 |
+
arxiv:
|
| 38 |
+
2510.14528
|
| 39 |
+
|
| 40 |
+
License:
|
| 41 |
+
apache-2.0
|
| 42 |
+
Model card
|
| 43 |
+
Files and versions
|
| 44 |
+
xet
|
| 45 |
+
Community
|
| 46 |
+
19
|
| 47 |
+
Settings
|
| 48 |
+
PaddleOCR-VL
|
| 49 |
+
/
|
| 50 |
+
README.md
|
| 51 |
+
|
| 52 |
+
ChengCui's picture
|
| 53 |
+
ChengCui
|
| 54 |
+
Update README.md
|
| 55 |
+
be4bd59
|
| 56 |
+
VERIFIED
|
| 57 |
+
about 22 hours ago
|
| 58 |
+
Preview
|
| 59 |
+
Code
|
| 60 |
+
|
|
| 61 |
+
raw
|
| 62 |
+
|
| 63 |
+
Copy download link
|
| 64 |
+
history
|
| 65 |
+
blame
|
| 66 |
+
|
| 67 |
+
20.5 kB
|
| 68 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
license: apache-2.0
|
| 70 |
pipeline_tag: image-text-to-text
|
| 71 |
tags:
|
|
|
|
| 78 |
- table
|
| 79 |
- formula
|
| 80 |
- chart
|
| 81 |
+
base_model: baidu/ERNIE-4.5-0.3B-Paddle
|
| 82 |
+
language:
|
| 83 |
+
- en
|
| 84 |
+
- zh
|
| 85 |
+
- multilingual
|
| 86 |
+
library_name: PaddleOCR
|
| 87 |
---
|
|
|
|
| 88 |
<div align="center">
|
| 89 |
|
| 90 |
|
|
|
|
| 94 |
|
| 95 |
</h1>
|
| 96 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 97 |
[](https://github.com/PaddlePaddle/PaddleOCR)
|
| 98 |
+
[](https://huggingface.co/PaddlePaddle/PaddleOCR-VL)
|
| 99 |
+
[](https://modelscope.cn/models/PaddlePaddle/PaddleOCR-VL)
|
| 100 |
+
[](https://huggingface.co/spaces/PaddlePaddle/PaddleOCR-VL_Online_Demo)
|
| 101 |
+
[](https://modelscope.cn/studios/PaddlePaddle/PaddleOCR-VL_Online_Demo/summary)
|
| 102 |
[](https://discord.gg/JPmZXDsEEK)
|
| 103 |
[](https://x.com/PaddlePaddle)
|
| 104 |
[](./LICENSE)
|
|
|
|
| 118 |
|
| 119 |
### **Core Features**
|
| 120 |
|
| 121 |
+
1. **Compact yet Powerful VLM Architecture:** We present a novel vision-language model that is specifically designed for resource-efficient inference, achieving outstanding performance in element recognition. By integrating a NaViT-style dynamic high-resolution visual encoder with the lightweight ERNIE-4.5-0.3B language model, we significantly enhance the model’s recognition capabilities and decoding efficiency. This integration maintains high accuracy while reducing computational demands, making it well-suited for efficient and practical document processing applications.
|
| 122 |
|
| 123 |
|
| 124 |
+
2. **SOTA Performance on Document Parsing:** PaddleOCR-VL achieves state-of-the-art performance in both page-level document parsing and element-level recognition. It significantly outperforms existing pipeline-based solutions and exhibiting strong competitiveness against leading vision-language models (VLMs) in document parsing. Moreover, it excels in recognizing complex document elements, such as text, tables, formulas, and charts, making it suitable for a wide range of challenging content types, including handwritten text and historical documents. This makes it highly versatile and suitable for a wide range of document types and scenarios.
|
| 125 |
|
| 126 |
|
| 127 |
+
3. **Multilingual Support:** PaddleOCR-VL Supports 109 languages, covering major global languages, including but not limited to Chinese, English, Japanese, Latin, and Korean, as well as languages with different scripts and structures, such as Russian (Cyrillic script), Arabic, Hindi (Devanagari script), and Thai. This broad language coverage substantially enhances the applicability of our system to multilingual and globalized document processing scenarios.
|
| 128 |
|
| 129 |
|
| 130 |
### **Model Architecture**
|
|
|
|
| 167 |
|
| 168 |
```python
|
| 169 |
from paddleocr import PaddleOCRVL
|
|
|
|
| 170 |
pipeline = PaddleOCRVL()
|
| 171 |
output = pipeline.predict("https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/paddleocr_vl_demo.png")
|
| 172 |
for res in output:
|
|
|
|
| 186 |
--network host \
|
| 187 |
ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddlex-genai-vllm-server
|
| 188 |
```
|
|
|
|
| 189 |
2. Call the PaddleOCR CLI or Python API:
|
| 190 |
|
| 191 |
```bash
|
|
|
|
| 194 |
--vl_rec_backend vllm-server \
|
| 195 |
--vl_rec_server_url http://127.0.0.1:8080/v1
|
| 196 |
```
|
|
|
|
| 197 |
```python
|
| 198 |
from paddleocr import PaddleOCRVL
|
|
|
|
| 199 |
pipeline = PaddleOCRVL(vl_rec_backend="vllm-server", vl_rec_server_url="http://127.0.0.1:8080/v1")
|
| 200 |
output = pipeline.predict("https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/paddleocr_vl_demo.png")
|
| 201 |
for res in output:
|
|
|
|
| 205 |
```
|
| 206 |
|
| 207 |
**For more usage details and parameter explanations, see the [documentation](https://www.paddleocr.ai/latest/en/version3.x/pipeline_usage/PaddleOCR-VL.html).**
|
|
|
|
| 208 |
## Performance
|
| 209 |
|
| 210 |
### Page-Level Document Parsing
|
|
|
|
| 357 |
primaryClass={cs.CV},
|
| 358 |
url={https://arxiv.org/abs/2510.14528},
|
| 359 |
}
|
| 360 |
+
```
|