Update README.md
Browse files
README.md
CHANGED
@@ -2,4 +2,44 @@
|
|
2 |
license: other
|
3 |
---
|
4 |
|
5 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: other
|
3 |
---
|
4 |
|
5 |
+
# PubTables-1M
|
6 |
+
|
7 |
+
- GitHub: [https://github.com/microsoft/table-transformer](https://github.com/microsoft/table-transformer)
|
8 |
+
- Paper: ["PubTables-1M: Towards comprehensive table extraction from unstructured documents"](https://openaccess.thecvf.com/content/CVPR2022/html/Smock_PubTables-1M_Towards_Comprehensive_Table_Extraction_From_Unstructured_Documents_CVPR_2022_paper.html)
|
9 |
+
- Hugging Face:
|
10 |
+
- [Detection model](https://huggingface.co/microsoft/table-transformer-detection)
|
11 |
+
- [Structure recognition model](https://huggingface.co/microsoft/table-transformer-structure-recognition)
|
12 |
+
|
13 |
+
Currently we only support downloading the dataset as tar.gz files. Integrating with HuggingFace Datasets is something we hope to support in the future!
|
14 |
+
|
15 |
+
Please switch to the "Files and versions" tab to download all of the files or use a command such as wget to download from the command line.
|
16 |
+
|
17 |
+
Once downloaded, use the included script "extract_structure_dataset.sh" to extract and organize all of the data.
|
18 |
+
|
19 |
+
## Files
|
20 |
+
|
21 |
+
It comes in 18 tar.gz files:
|
22 |
+
- Training and evaluation data for the structure recognition model (947,642 total cropped table instances):
|
23 |
+
- PubTables-1M-Structure_Filelists.tar.gz
|
24 |
+
- PubTables-1M-Structure_Annotations_Test.tar.gz: 93,834 XML files containing bounding boxes in PASCAL VOC format
|
25 |
+
- PubTables-1M-Structure_Annotations_Train.tar.gz: 758,849 XML files containing bounding boxes in PASCAL VOC format
|
26 |
+
- PubTables-1M-Structure_Annotations_Val.tar.gz: 94,959 XML files containing bounding boxes in PASCAL VOC format
|
27 |
+
- PubTables-1M-Structure_Images_Test.tar.gz
|
28 |
+
- PubTables-1M-Structure_Images_Train.tar.gz
|
29 |
+
- PubTables-1M-Structure_Images_Val.tar.gz
|
30 |
+
- PubTables-1M-Structure_Table_Words.tar.gz: Bounding boxes and text content for all of the words in each cropped table image
|
31 |
+
|
32 |
+
Training and evaluation data for the detection model (575,305 total document page instances):
|
33 |
+
- PubTables-1M-Detection_Filelists.tar.gz
|
34 |
+
- PubTables-1M-Detection_Annotations_Test.tar.gz: 57,125 XML files containing bounding boxes in PASCAL VOC format
|
35 |
+
- PubTables-1M-Detection_Annotations_Train.tar.gz: 460,589 XML files containing bounding boxes in PASCAL VOC format
|
36 |
+
- PubTables-1M-Detection_Annotations_Val.tar.gz: 57,591 XML files containing bounding boxes in PASCAL VOC format
|
37 |
+
- PubTables-1M-Detection_Images_Test.tar.gz
|
38 |
+
- PubTables-1M-Detection_Images_Train_Part1.tar.gz
|
39 |
+
- PubTables-1M-Detection_Images_Train_Part2.tar.gz
|
40 |
+
- PubTables-1M-Detection_Images_Val.tar.gz
|
41 |
+
- PubTables-1M-Detection_Page_Words.tar.gz: Bounding boxes and text content for all of the words in each page image (plus some unused files)
|
42 |
+
|
43 |
+
Full table annotations for the source PDF files:
|
44 |
+
- PubTables-1M-PDF_Annotations.tar.gz: Detailed annotations for all of the tables appearing in the source PubMed PDFs. All annotations are in PDF coordinates.
|
45 |
+
- 401,733 JSON files, one per source PDF document
|