Views
No views yet
torch==2.0.1
torchvision==0.15.2
transformers==4.37.2
tiktoken==0.6.0
verovio==4.3.1
accelerate==0.28.01from transformers import AutoModel, AutoTokenizer
2
3tokenizer = AutoTokenizer.from_pretrained('ucaslcl/GOT-OCR2_0', trust_remote_code=True)
4model = AutoModel.from_pretrained('ucaslcl/GOT-OCR2_0', trust_remote_code=True, low_cpu_mem_usage=True, device_map='cuda', use_safetensors=True, pad_token_id=tokenizer.eos_token_id)
5model = model.eval().cuda()
6
7
8# input your test image
9image_file = 'xxx.jpg'
10
11# plain texts OCR
12res = model.chat(tokenizer, image_file, ocr_type='ocr')
13
14# format texts OCR:
15# res = model.chat(tokenizer, image_file, ocr_type='format')
16
17# fine-grained OCR:
18# res = model.chat(tokenizer, image_file, ocr_type='ocr', ocr_box='')
19# res = model.chat(tokenizer, image_file, ocr_type='format', ocr_box='')
20# res = model.chat(tokenizer, image_file, ocr_type='ocr', ocr_color='')
21# res = model.chat(tokenizer, image_file, ocr_type='format', ocr_color='')
22
23# multi-crop OCR:
24# res = model.chat_crop(tokenizer, image_file, ocr_type='ocr')
25# res = model.chat_crop(tokenizer, image_file, ocr_type='format')
26
27# render the formatted OCR results:
28# res = model.chat(tokenizer, image_file, ocr_type='format', render=True, save_render_file = './demo.html')
29
30print(res)
31
321@article{wei2024general,
2 title={General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model},
3 author={Wei, Haoran and Liu, Chenglong and Chen, Jinyue and Wang, Jia and Kong, Lingyu and Xu, Yanming and Ge, Zheng and Zhao, Liang and Sun, Jianjian and Peng, Yuang and others},
4 journal={arXiv preprint arXiv:2409.01704},
5 year={2024}
6}
7@article{liu2024focus,
8 title={Focus Anywhere for Fine-grained Multi-page Document Understanding},
9 author={Liu, Chenglong and Wei, Haoran and Chen, Jinyue and Kong, Lingyu and Ge, Zheng and Zhu, Zining and Zhao, Liang and Sun, Jianjian and Han, Chunrui and Zhang, Xiangyu},
10 journal={arXiv preprint arXiv:2405.14295},
11 year={2024}
12}
13@article{wei2023vary,
14 title={Vary: Scaling up the Vision Vocabulary for Large Vision-Language Models},
15 author={Wei, Haoran and Kong, Lingyu and Chen, Jinyue and Zhao, Liang and Ge, Zheng and Yang, Jinrong and Sun, Jianjian and Han, Chunrui and Zhang, Xiangyu},
16 journal={arXiv preprint arXiv:2312.06109},
17 year={2023}
18}