The models trained in our baseline are listed below. All models were trained 10 epochs under their respective backbones. Using the pretrained model provided by Detectron2.
The models trained on TableBank are available in the format used by Detectron2.
| Models | Url | Size |
|---|---|---|
| X101(Word) | Model/Config | 796MB |
| X152(Word) | Model/Config | 1.03GB |
| X101(Latex) | Model/Config | 796MB |
| X152(Latex) | Model/Config | 1.03GB |
| X101(Word+Latex) | Model/Config | 796MB |
| X152(Word+Latex) | Model/Config | 1.03GB |
The models trained on TableBank are available in the format used by OpenNMT.
| Models | Url | Size |
|---|---|---|
| Image-to-Text (Word) | Model | 49.5MB |
| Image-to-Text (Latex) | Model | 49.5MB |
| Image-to-Text (Word+Latex) | Model | 49.5MB |