|
| 1 | +<!--Copyright 2024 StepFun and The HuggingFace Team. All rights reserved. |
| 2 | +
|
| 3 | +Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with |
| 4 | +the License. You may obtain a copy of the License at |
| 5 | +
|
| 6 | +http://www.apache.org/licenses/LICENSE-2.0 |
| 7 | +
|
| 8 | +Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on |
| 9 | +an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the |
| 10 | +specific language governing permissions and limitations under the License. |
| 11 | +
|
| 12 | +⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be |
| 13 | +rendered properly in your Markdown viewer. |
| 14 | +
|
| 15 | +--> |
| 16 | + |
| 17 | +# GOT-OCR2 |
| 18 | + |
| 19 | +## Overview |
| 20 | + |
| 21 | +The GOT-OCR2 model was proposed in [General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model](https://arxiv.org/abs/2409.01704) by Haoran Wei, Chenglong Liu, Jinyue Chen, Jia Wang, Lingyu Kong, Yanming Xu, Zheng Ge, Liang Zhao, Jianjian Sun, Yuang Peng, Chunrui Han, Xiangyu Zhang. |
| 22 | + |
| 23 | +The abstract from the paper is the following: |
| 24 | + |
| 25 | +*Traditional OCR systems (OCR-1.0) are increasingly unable to meet people’snusage due to the growing demand for intelligent processing of man-made opticalncharacters. In this paper, we collectively refer to all artificial optical signals (e.g., plain texts, math/molecular formulas, tables, charts, sheet music, and even geometric shapes) as "characters" and propose the General OCR Theory along with an excellent model, namely GOT, to promote the arrival of OCR-2.0. The GOT, with 580M parameters, is a unified, elegant, and end-to-end model, consisting of a high-compression encoder and a long-contexts decoder. As an OCR-2.0 model, GOT can handle all the above "characters" under various OCR tasks. On the input side, the model supports commonly used scene- and document-style images in slice and whole-page styles. On the output side, GOT can generate plain or formatted results (markdown/tikz/smiles/kern) via an easy prompt. Besides, the model enjoys interactive OCR features, i.e., region-level recognition guided by coordinates or colors. Furthermore, we also adapt dynamic resolution and multipage OCR technologies to GOT for better practicality. In experiments, we provide sufficient results to prove the superiority of our model.* |
| 26 | + |
| 27 | +<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/got_ocr_overview.png" |
| 28 | +alt="drawing" width="600"/> |
| 29 | + |
| 30 | +<small> GOT-OCR2 training stages. Taken from the <a href="https://arxiv.org/abs/2409.01704">original paper.</a> </small> |
| 31 | + |
| 32 | + |
| 33 | +Tips: |
| 34 | + |
| 35 | +GOT-OCR2 works on a wide range of tasks, including plain document OCR, scene text OCR, formatted document OCR, and even OCR for tables, charts, mathematical formulas, geometric shapes, molecular formulas and sheet music. While this implementation of the model will only output plain text, the outputs can be further processed to render the desired format, with packages like `pdftex`, `mathpix`, `matplotlib`, `tikz`, `verovio` or `pyecharts`. |
| 36 | +The model can also be used for interactive OCR, where the user can specify the region to be recognized by providing the coordinates or the color of the region's bounding box. |
| 37 | + |
| 38 | +This model was contributed by [yonigozlan](https://huggingface.co/yonigozlan). |
| 39 | +The original code can be found [here](https://github.com/Ucas-HaoranWei/GOT-OCR2.0). |
| 40 | + |
| 41 | +## Usage example |
| 42 | + |
| 43 | +### Plain text inference |
| 44 | + |
| 45 | +```python |
| 46 | +>>> from transformers import AutoProcessor, AutoModelForImageTextToText |
| 47 | + |
| 48 | +>>> device = "cuda" if torch.cuda.is_available() else "cpu" |
| 49 | +>>> model = AutoModelForImageTextToText.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf", device_map=device) |
| 50 | +>>> processor = AutoProcessor.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf") |
| 51 | + |
| 52 | +>>> image = "https://huggingface.co/datasets/hf-internal-testing/fixtures_got_ocr/resolve/main/image_ocr.jpg" |
| 53 | +>>> inputs = processor(image, return_tensors="pt").to(device) |
| 54 | + |
| 55 | +>>> generate_ids = model.generate( |
| 56 | +... **inputs, |
| 57 | +... do_sample=False, |
| 58 | +... tokenizer=processor.tokenizer, |
| 59 | +... stop_strings="<|im_end|>", |
| 60 | +... max_new_tokens=4096, |
| 61 | +... ) |
| 62 | + |
| 63 | +>>> processor.decode(generate_ids[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True) |
| 64 | +"R&D QUALITY IMPROVEMENT\nSUGGESTION/SOLUTION FORM\nName/Phone Ext. : (...)" |
| 65 | +``` |
| 66 | + |
| 67 | +### Plain text inference batched |
| 68 | + |
| 69 | +```python |
| 70 | +>>> from transformers import AutoProcessor, AutoModelForImageTextToText |
| 71 | + |
| 72 | +>>> device = "cuda" if torch.cuda.is_available() else "cpu" |
| 73 | +>>> model = AutoModelForImageTextToText.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf", device_map=device) |
| 74 | +>>> processor = AutoProcessor.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf") |
| 75 | + |
| 76 | +>>> image1 = "https://huggingface.co/datasets/hf-internal-testing/fixtures_got_ocr/resolve/main/multi_box.png" |
| 77 | +>>> image2 = "https://huggingface.co/datasets/hf-internal-testing/fixtures_got_ocr/resolve/main/image_ocr.jpg" |
| 78 | + |
| 79 | +>>> inputs = processor([image1, image2], return_tensors="pt").to(device) |
| 80 | + |
| 81 | +>>> generate_ids = model.generate( |
| 82 | +... **inputs, |
| 83 | +... do_sample=False, |
| 84 | +... tokenizer=processor.tokenizer, |
| 85 | +... stop_strings="<|im_end|>", |
| 86 | +... max_new_tokens=4, |
| 87 | +... ) |
| 88 | + |
| 89 | +>>> processor.batch_decode(generate_ids[:, inputs["input_ids"].shape[1] :], skip_special_tokens=True) |
| 90 | +["Reducing the number", "R&D QUALITY"] |
| 91 | +``` |
| 92 | + |
| 93 | +### Formatted text inference |
| 94 | + |
| 95 | +GOT-OCR2 can also generate formatted text, such as markdown or LaTeX. Here is an example of how to generate formatted text: |
| 96 | + |
| 97 | +```python |
| 98 | +>>> from transformers import AutoProcessor, AutoModelForImageTextToText |
| 99 | + |
| 100 | +>>> device = "cuda" if torch.cuda.is_available() else "cpu" |
| 101 | +>>> model = AutoModelForImageTextToText.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf", device_map=device) |
| 102 | +>>> processor = AutoProcessor.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf") |
| 103 | + |
| 104 | +>>> image = "https://huggingface.co/datasets/hf-internal-testing/fixtures_got_ocr/resolve/main/latex.png" |
| 105 | +>>> inputs = processor(image, return_tensors="pt", format=True).to(device) |
| 106 | + |
| 107 | +>>> generate_ids = model.generate( |
| 108 | +... **inputs, |
| 109 | +... do_sample=False, |
| 110 | +... tokenizer=processor.tokenizer, |
| 111 | +... stop_strings="<|im_end|>", |
| 112 | +... max_new_tokens=4096, |
| 113 | +... ) |
| 114 | + |
| 115 | +>>> processor.decode(generate_ids[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True) |
| 116 | +"\\author{\nHanwen Jiang* \\(\\quad\\) Arjun Karpur \\({ }^{\\dagger} \\quad\\) Bingyi Cao \\({ }^{\\dagger} \\quad\\) (...)" |
| 117 | +``` |
| 118 | + |
| 119 | +### Inference on multiple pages |
| 120 | + |
| 121 | +Although it might be reasonable in most cases to use a “for loop” for multi-page processing, some text data with formatting across several pages make it necessary to process all pages at once. GOT introduces a multi-page OCR (without “for loop”) feature, where multiple pages can be processed by the model at once, whith the output being one continuous text. |
| 122 | +Here is an example of how to process multiple pages at once: |
| 123 | + |
| 124 | + |
| 125 | +```python |
| 126 | +>>> from transformers import AutoProcessor, AutoModelForImageTextToText |
| 127 | + |
| 128 | +>>> device = "cuda" if torch.cuda.is_available() else "cpu" |
| 129 | +>>> model = AutoModelForImageTextToText.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf", device_map=device) |
| 130 | +>>> processor = AutoProcessor.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf") |
| 131 | + |
| 132 | +>>> image1 = "https://huggingface.co/datasets/hf-internal-testing/fixtures_got_ocr/resolve/main/page1.png" |
| 133 | +>>> image2 = "https://huggingface.co/datasets/hf-internal-testing/fixtures_got_ocr/resolve/main/page2.png" |
| 134 | +>>> inputs = processor([image1, image2], return_tensors="pt", multi_page=True, format=True).to(device) |
| 135 | + |
| 136 | +>>> generate_ids = model.generate( |
| 137 | +... **inputs, |
| 138 | +... do_sample=False, |
| 139 | +... tokenizer=processor.tokenizer, |
| 140 | +... stop_strings="<|im_end|>", |
| 141 | +... max_new_tokens=4096, |
| 142 | +... ) |
| 143 | + |
| 144 | +>>> processor.decode(generate_ids[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True) |
| 145 | +"\\title{\nGeneral OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model\n}\n\\author{\nHaoran Wei (...)" |
| 146 | +``` |
| 147 | + |
| 148 | +### Inference on cropped patches |
| 149 | + |
| 150 | +GOT supports a 1024×1024 input resolution, which is sufficient for most OCR tasks, such as scene OCR or processing A4-sized PDF pages. However, certain scenarios, like horizontally stitched two-page PDFs commonly found in academic papers or images with unusual aspect ratios, can lead to accuracy issues when processed as a single image. To address this, GOT can dynamically crop an image into patches, process them all at once, and merge the results for better accuracy with such inputs. |
| 151 | +Here is an example of how to process cropped patches: |
| 152 | + |
| 153 | +```python |
| 154 | +>>> import torch |
| 155 | +>>> from transformers import AutoProcessor, AutoModelForImageTextToText |
| 156 | + |
| 157 | +>>> device = "cuda" if torch.cuda.is_available() else "cpu" |
| 158 | +>>> model = AutoModelForImageTextToText.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf", torch_dtype=torch.bfloat16, device_map=device) |
| 159 | +>>> processor = AutoProcessor.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf") |
| 160 | + |
| 161 | +>>> image = "https://huggingface.co/datasets/hf-internal-testing/fixtures_got_ocr/resolve/main/one_column.png" |
| 162 | +>>> inputs = processor(image, return_tensors="pt", format=True, crop_to_patches=True, max_patches=3).to(device) |
| 163 | + |
| 164 | +>>> generate_ids = model.generate( |
| 165 | +... **inputs, |
| 166 | +... do_sample=False, |
| 167 | +... tokenizer=processor.tokenizer, |
| 168 | +... stop_strings="<|im_end|>", |
| 169 | +... max_new_tokens=4096, |
| 170 | +... ) |
| 171 | + |
| 172 | +>>> processor.decode(generate_ids[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True) |
| 173 | +"on developing architectural improvements to make learnable matching methods generalize.\nMotivated by the above observations, (...)" |
| 174 | +``` |
| 175 | + |
| 176 | +### Inference on a specific region |
| 177 | + |
| 178 | +GOT supports interactive OCR, where the user can specify the region to be recognized by providing the coordinates or the color of the region's bounding box. Here is an example of how to process a specific region: |
| 179 | + |
| 180 | +```python |
| 181 | +>>> from transformers import AutoProcessor, AutoModelForImageTextToText |
| 182 | + |
| 183 | +>>> device = "cuda" if torch.cuda.is_available() else "cpu" |
| 184 | +>>> model = AutoModelForImageTextToText.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf", device_map=device) |
| 185 | +>>> processor = AutoProcessor.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf") |
| 186 | + |
| 187 | +>>> image = "https://huggingface.co/datasets/hf-internal-testing/fixtures_got_ocr/resolve/main/multi_box.png" |
| 188 | +>>> inputs = processor(image, return_tensors="pt", color="green").to(device) # or box=[x1, y1, x2, y2] for coordinates (image pixels) |
| 189 | + |
| 190 | +>>> generate_ids = model.generate( |
| 191 | +... **inputs, |
| 192 | +... do_sample=False, |
| 193 | +... tokenizer=processor.tokenizer, |
| 194 | +... stop_strings="<|im_end|>", |
| 195 | +... max_new_tokens=4096, |
| 196 | +... ) |
| 197 | + |
| 198 | +>>> processor.decode(generate_ids[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True) |
| 199 | +"You should keep in mind what features from the module should be used, especially \nwhen you’re planning to sell a template." |
| 200 | +``` |
| 201 | + |
| 202 | +### Inference on general OCR data example: sheet music |
| 203 | + |
| 204 | +Although this implementation of the model will only output plain text, the outputs can be further processed to render the desired format, with packages like `pdftex`, `mathpix`, `matplotlib`, `tikz`, `verovio` or `pyecharts`. |
| 205 | +Here is an example of how to process sheet music: |
| 206 | + |
| 207 | +```python |
| 208 | +>>> from transformers import AutoProcessor, AutoModelForImageTextToText |
| 209 | +>>> import verovio |
| 210 | + |
| 211 | +>>> device = "cuda" if torch.cuda.is_available() else "cpu" |
| 212 | +>>> model = AutoModelForImageTextToText.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf", device_map=device) |
| 213 | +>>> processor = AutoProcessor.from_pretrained("stepfun-ai/GOT-OCR-2.0-hf") |
| 214 | + |
| 215 | +>>> image = "https://huggingface.co/datasets/hf-internal-testing/fixtures_got_ocr/resolve/main/sheet_music.png" |
| 216 | +>>> inputs = processor(image, return_tensors="pt", format=True).to(device) |
| 217 | + |
| 218 | +>>> generate_ids = model.generate( |
| 219 | +... **inputs, |
| 220 | +... do_sample=False, |
| 221 | +... tokenizer=processor.tokenizer, |
| 222 | +... stop_strings="<|im_end|>", |
| 223 | +... max_new_tokens=4096, |
| 224 | +... ) |
| 225 | + |
| 226 | +>>> outputs = processor.decode(generate_ids[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True) |
| 227 | +>>> tk = verovio.toolkit() |
| 228 | +>>> tk.loadData(outputs) |
| 229 | +>>> tk.setOptions( |
| 230 | +... { |
| 231 | +... "pageWidth": 2100, |
| 232 | +... "pageHeight": 800, |
| 233 | +... "footer": "none", |
| 234 | +... "barLineWidth": 0.5, |
| 235 | +... "beamMaxSlope": 15, |
| 236 | +... "staffLineWidth": 0.2, |
| 237 | +... "spacingStaff": 6, |
| 238 | +... } |
| 239 | +... ) |
| 240 | +>>> tk.getPageCount() |
| 241 | +>>> svg = tk.renderToSVG() |
| 242 | +>>> svg = svg.replace('overflow="inherit"', 'overflow="visible"') |
| 243 | +>>> with open("output.svg", "w") as f: |
| 244 | +>>> f.write(svg) |
| 245 | +``` |
| 246 | +<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/sheet_music.svg" |
| 247 | +alt="drawing" width="600"/> |
| 248 | + |
| 249 | +## GotOcr2Config |
| 250 | + |
| 251 | +[[autodoc]] GotOcr2Config |
| 252 | + |
| 253 | +## GotOcr2VisionConfig |
| 254 | + |
| 255 | +[[autodoc]] GotOcr2VisionConfig |
| 256 | + |
| 257 | +## GotOcr2ImageProcessor |
| 258 | + |
| 259 | +[[autodoc]] GotOcr2ImageProcessor |
| 260 | + |
| 261 | +## GotOcr2Processor |
| 262 | + |
| 263 | +[[autodoc]] GotOcr2Processor |
| 264 | + |
| 265 | +## GotOcr2ForConditionalGeneration |
| 266 | + |
| 267 | +[[autodoc]] GotOcr2ForConditionalGeneration |
| 268 | + - forward |
| 269 | + |
0 commit comments