hf-tuner/donut-base-finetuned-rvl-cdip
Image-Text-to-Text
•
0.2B
•
Updated
•
103
image
imagewidth (px) 607
1.47k
| label
class label 16
classes |
|---|---|
2email
|
|
4advertisement
|
|
3handwritten
|
|
11invoice
|
|
15memo
|
|
13questionnaire
|
|
6scientific publication
|
|
3handwritten
|
|
8file folder
|
|
8file folder
|
|
12presentation
|
|
12presentation
|
|
13questionnaire
|
|
1form
|
|
4advertisement
|
|
1form
|
|
9news article
|
|
12presentation
|
|
15memo
|
|
0letter
|
|
5scientific report
|
|
12presentation
|
|
12presentation
|
|
4advertisement
|
|
6scientific publication
|
|
2email
|
|
15memo
|
|
10budget
|
|
13questionnaire
|
|
8file folder
|
|
3handwritten
|
|
13questionnaire
|
|
11invoice
|
|
11invoice
|
|
10budget
|
|
6scientific publication
|
|
0letter
|
|
3handwritten
|
|
15memo
|
|
13questionnaire
|
|
8file folder
|
|
3handwritten
|
|
6scientific publication
|
|
14resume
|
|
12presentation
|
|
14resume
|
|
12presentation
|
|
2email
|
|
11invoice
|
|
6scientific publication
|
|
13questionnaire
|
|
2email
|
|
13questionnaire
|
|
4advertisement
|
|
3handwritten
|
|
7specification
|
|
8file folder
|
|
14resume
|
|
6scientific publication
|
|
4advertisement
|
|
3handwritten
|
|
8file folder
|
|
15memo
|
|
13questionnaire
|
|
8file folder
|
|
5scientific report
|
|
3handwritten
|
|
6scientific publication
|
|
9news article
|
|
5scientific report
|
|
9news article
|
|
5scientific report
|
|
0letter
|
|
2email
|
|
14resume
|
|
4advertisement
|
|
1form
|
|
11invoice
|
|
12presentation
|
|
9news article
|
|
9news article
|
|
8file folder
|
|
2email
|
|
10budget
|
|
11invoice
|
|
14resume
|
|
11invoice
|
|
7specification
|
|
14resume
|
|
2email
|
|
1form
|
|
14resume
|
|
15memo
|
|
4advertisement
|
|
1form
|
|
3handwritten
|
|
2email
|
|
12presentation
|
|
4advertisement
|
|
4advertisement
|
This dataset is created from original aharley/rvl_cdip dataset using this notebook
This dataset consists of 8992 grayscale images in 16 classes, with 562 images per class. There are 8000 training images(500 image per class) and 992 test images(62 images per class). The images are sized so their largest dimension does not exceed 1000 pixels.