| import os | |
| import gradio as gr | |
| import numpy as np | |
| import layoutparser as lp | |
| from PIL import Image | |
| import PIL | |
| os.system('pip install "git+https://github.com/facebookresearch/[email protected]#egg=detectron2" ') | |
| model = lp.Detectron2LayoutModel('lp://PubLayNet/faster_rcnn_R_50_FPN_3x/config', | |
| extra_config=["MODEL.ROI_HEADS.SCORE_THRESH_TEST", 0.8], | |
| label_map={0: "Text", 1: "Title", 2: "List", 3:"Table", 4:"Figure"}) | |
| article="References<br>[1] Z. Shen, R. Zhang, M. Dell, B. C. G. Lee, J. Carlson, and W. Li, “LayoutParser: A Unified Toolkit for Deep Learning Based Document Image Analysis,” arXiv Prepr. arXiv2103.15348, 2021." | |
| description = "Layout Detection/Parsing is one of the important tasks of converting unstructured data into structured data. This task helps to automate, digitize and organize the data in a usable format. In this project, we utilize LayoutParser library (https://github.com/Layout-Parser/layout-parser) to perform Layout Detection using pre-trained Faster_rcnn_R_50_FPN model that can classify the layout based on Text, Title, List, Table and Figure. Upload an image of a document or click an example image to check this out." | |
| def show_preds(input_image): | |
| img = PIL.Image.fromarray(input_image, 'RGB') | |
| basewidth = 900 | |
| wpercent = (basewidth/float(img.size[0])) | |
| hsize = int((float(img.size[1])*float(wpercent))) | |
| img = img.resize((basewidth,hsize), Image.ANTIALIAS) | |
| image_array=np.array(img) | |
| layout = model.detect(image_array) | |
| return lp.draw_box(image_array, layout, show_element_type=True) | |
| outputs = gr.outputs.Image(type="pil") | |
| examples = [['example1.png'], ['example2.png']] | |
| gr_interface = gr.Interface(fn=show_preds, inputs=["image"], outputs=outputs, title='Document Layout Detector/Parser', article=article, description=description, examples=examples, analytics_enabled = True, enable_queue=True) | |
| gr_interface.launch(inline=False, share=True, debug=True) |