search

microsoft / table-transformer

Table Transformer (TATR) is a deep learning model for extracting tables from unstructured documents (PDFs and images). This is also the official repository for the PubTables-1M dataset and GriTS evaluation metric.
MIT License
2.01k stars 231 forks source link

What does spans mean in the cell output? #170

Open dipikakhullar opened 4 months ago

dipikakhullar commented 4 months ago

Hi, these are some example cells from the TATR recognize model output: 35.420074462890625, 474.67864990234375, 47.1126594543457], 'column_nums': [5], 'row_nums': [1], 'column header': True, 'subcell': False, 'projected row header': False, 'cell text': 'max', 'spans': [], 'bbox_input': [501.5734062194824, 281.6884765625, 527.1918754577637, 293.3810615539551]},

{'bbox': [242.34381103515625, 10.655202865600586, 336.2829284667969, 47.1126594543457], 'column_nums': [3], 'row_nums': [0, 1], 'column header': True, 'projected row header': False, 'cell text': 'Max. Thickness (T) mm/inches', 'spans': [], 'bbox_input': [294.8570365905762, 256.92360496520996, 388.7961540222168, 293.3810615539551]},

I have a few questions:

  1. what does "spans" mean? When is spans non empty? Any example?
  2. Is there a concept of row spans and column spans?
  3. what is 'projected row header'?
  4. What is 'column header'? And how does this differ from 'projected row header'?