PDF Extraction

This project focuses on extracting parameters from PDF files, specifically by recognizing tables. The main tasks include:

Table Extraction

The table extraction process utilizes the Cycle-CenterNet model for table structure recognition. The model is designed to identify structured tables in images and convert them into a format that a large language model can understand.

Model Setup

1. Clone the Model: First, install Git LFS and clone the table structure recognition model repository:

   git lfs install
   git clone https://www.modelscope.cn/iic/cv_dla34_table-structure-recognition_cycle-centernet.git

2. Model Path: After cloning the repository, set the absolute path to the model as follows:

   model_path = "your_absolute_path_to_cv_dla34_table-structure-recognition_cycle-centernet"

OCR Setup

To extract text from images, we use PaddleOCR. PaddleOCR is capable of recognizing text in different languages and works in conjunction with the table structure model.

OCR Setup

Set the absolute paths for the OCR models:

from paddleocr import PaddleOCR

ocr = PaddleOCR(
    use_gpu=True,
    lang='ch',  # Set language to Chinese ('ch')
    det_model_dir='your_absolute_path_to_ch_PP-OCRv4_det_infer',
    rec_model_dir='your_absolute_path_to_ch_PP-OCRv4_rec_infer',
    cls_model_dir='your_absolute_path_to_ch_ppocr_mobile_v2.0_cls_infer'
)

After starting the code, the PaddleOCR model will be automatically downloaded:

python main.py

or

python Table_Extraction.py

Effect display

PartNumber | TotalCapacitance(Ct) @ 50 V,(pF) | TotalCapacitance(Ct) @ 0 V,(pF) | SeriesResistance (Rs),@10 mA,(②) | MinorityCarrierLifetime (TL)@ 10 mA(ns) | VoltageRating2(M) | I-RegionThickness(μm) | ThermalResistance(0JC)(°C/W)
Maximum | Typical | Maximum | Typical | Minimum | Nominal | Maximum

Switching Applications(continued)
 | APD0810-203 | 0.35 | 0.40 | 1.5 | 160 | 100 | 8 | 174
 | APD0810-210 | 0.40 | 0.45 | 1.5 | 160 | 100 | 8 | 75
 | APD0810-219 | 0.35 | 0.40 | 1.5 | 160 | 100 | 8 | 143
 | APD0810-240 | 0.35 | 0.40 | 1.5 | 160 | 100 | 8 | 155
 | APD1505-203 | 0.40 | 0.45@10V | 2.5 | 350 | 200 | 15 | 172
 | APD1505-210 | 0.40 | 0.45@10 V | 2.5 | 350 | 200 | 15 | 74
 | APD1505-219 | 0.40 | 0.45@10V | 2.5 | 350 | 200 | 15 | 142
 | APD1505-240 | 0.40 | 0.45 @ 10 V | 2.5 | 350 | 200 | 15 | 150
 | APD1510-203 | 0.35 | 0.40 | 2.0 | 300 | 200 | 15 | 168
 | APD1510-210 | 0.35 | 0.40 | 2.0 | 300 | 200 | 15 | 70
 | APD1510-219 | 0.35 | 0.40 | 2.0 | 300 | 200 | 15 | 137
 | APD1510-240 | 0.35 | 0.40 | 2.0 | 300 | 200 | 15 | 149
 | APD1520-203 | 0.40 | 0.45 | 1.2 | 900 | 200 | 15 | 155
 | APD1520-210 | 0.40 | 0.45 | 1.2 | 900 | 200 | 15 | 57
 | APD1520-219 | 0.45 | 0.50 | 1.2 | 900 | 200 | 15 | 124
 | APD1520-240 | 0.40 | 0.45 | 1.2 | 900 | 200 | 15 | 136

AttenuatorApplications
 | APD2220-203 | 0.45 | 0.50 | 4.0 | 100 | 100 | 50 | 132
 | APD2220-210 | 0.45 | 0.50 | 4.0 | 100 | 100 | 50 | 32
 | APD2220-219 | 0.40 | 0.45 | 4.0 | 100 | 100 | 50 | 104
 | APD2220-240 | 0.40 | 0.45 | 4.0 | 100 | 100 | 50 | 115

Usage

Once the models are set up, you can start extracting tables and text from images.

This project makes use of wyf3