Spaces:
Runtime error
Runtime error
| import os | |
| os.system('pip install pip --upgrade') | |
| os.system('pip install -q git+https://github.com/huggingface/transformers.git') | |
| os.system("pip install pyyaml==5.1") | |
| # workaround: install old version of pytorch since detectron2 hasn't released packages for pytorch 1.9 (issue: https://github.com/facebookresearch/detectron2/issues/3158) | |
| os.system( | |
| "pip install torch==1.8.0+cu101 torchvision==0.9.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html" | |
| ) | |
| # install detectron2 that matches pytorch 1.8 | |
| # See https://detectron2.readthedocs.io/tutorials/install.html for instructions | |
| os.system( | |
| "pip install -q detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu101/torch1.8/index.html" | |
| ) | |
| ## install PyTesseract | |
| os.system("pip install -q pytesseract") | |
| import gradio as gr | |
| import numpy as np | |
| from transformers import LayoutLMv3Processor, LayoutLMv3ForTokenClassification | |
| from datasets import load_dataset | |
| from PIL import Image, ImageDraw, ImageFont | |
| processor = LayoutLMv3Processor.from_pretrained("microsoft/layoutlmv3-base") | |
| model = LayoutLMv3ForTokenClassification.from_pretrained( | |
| "jinhybr/OCR-LayoutLMv3" | |
| ) | |
| # load image example | |
| dataset = load_dataset("nielsr/funsd", split="test") | |
| image = Image.open(dataset[0]["image_path"]).convert("RGB") | |
| image = Image.open("./example_lm3.png") | |
| image.save("document.png") | |
| labels = dataset.features["ner_tags"].feature.names | |
| id2label = {v: k for v, k in enumerate(labels)} | |
| label2color = { | |
| "question": "blue", | |
| "answer": "green", | |
| "header": "orange", | |
| "other": "violet", | |
| } | |
| def unnormalize_box(bbox, width, height): | |
| return [ | |
| width * (bbox[0] / 1000), | |
| height * (bbox[1] / 1000), | |
| width * (bbox[2] / 1000), | |
| height * (bbox[3] / 1000), | |
| ] | |
| def iob_to_label(label): | |
| label = label[2:] | |
| if not label: | |
| return "other" | |
| return label | |
| def process_image(image): | |
| width, height = image.size | |
| # encode | |
| encoding = processor( | |
| image, truncation=True, return_offsets_mapping=True, return_tensors="pt" | |
| ) | |
| offset_mapping = encoding.pop("offset_mapping") | |
| # forward pass | |
| outputs = model(**encoding) | |
| # get predictions | |
| predictions = outputs.logits.argmax(-1).squeeze().tolist() | |
| token_boxes = encoding.bbox.squeeze().tolist() | |
| # only keep non-subword predictions | |
| is_subword = np.array(offset_mapping.squeeze().tolist())[:, 0] != 0 | |
| true_predictions = [ | |
| id2label[pred] for idx, pred in enumerate(predictions) if not is_subword[idx] | |
| ] | |
| true_boxes = [ | |
| unnormalize_box(box, width, height) | |
| for idx, box in enumerate(token_boxes) | |
| if not is_subword[idx] | |
| ] | |
| # draw predictions over the image | |
| draw = ImageDraw.Draw(image) | |
| font = ImageFont.load_default() | |
| for prediction, box in zip(true_predictions, true_boxes): | |
| predicted_label = iob_to_label(prediction).lower() | |
| draw.rectangle(box, outline=label2color[predicted_label]) | |
| draw.text( | |
| (box[0] + 10, box[1] - 10), | |
| text=predicted_label, | |
| fill=label2color[predicted_label], | |
| font=font, | |
| ) | |
| return image | |
| title = "OCR Document Parser : Information Extraction - Fine Tuned LayoutLMv3 Model" | |
| description = "Demo for Microsoft's LayoutLMv3, a Transformer for state-of-the-art document image understanding tasks. This particular model is fine-tuned on FUNSD, a dataset of manually annotated forms. It annotates the words appearing in the image as QUESTION/ANSWER/HEADER/OTHER. To use it, simply upload an image or use the example image below and click 'Submit'. Results will show up in a few seconds. If you want to make the output bigger, right-click on it and select 'Open image in new tab'." | |
| article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2204.08387' target='_blank'>LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking</a> | <a href='https://github.com/microsoft/unilm' target='_blank'>Github Repo</a></p>" | |
| examples = [["document.png"]] | |
| css = ".output-image, .input-image {height: 40rem !important; width: 100% !important;}" | |
| # css = "@media screen and (max-width: 600px) { .output_image, .input_image {height:20rem !important; width: 100% !important;} }" | |
| # css = ".output_image, .input_image {height: 600px !important}" | |
| css = ".image-preview {height: auto !important;}" | |
| iface = gr.Interface( | |
| fn=process_image, | |
| inputs=gr.inputs.Image(type="pil"), | |
| outputs=gr.outputs.Image(type="pil", label="annotated image"), | |
| title=title, | |
| description=description, | |
| article=article, | |
| examples=examples, | |
| css=css, | |
| enable_queue=True, | |
| ) | |
| iface.launch(debug=True) |