You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, first of all, thank you so much for releasing such a good source.
It works pretty well on our dataset, but we got incorrect layout detection in a few cases and want to further fine-tune the model with the layoutlmv3-sft pretrained model.
So, I would like to understand the format of the datasets used in the actual layout detection, could you please provide a link to the dataset source or related materials?
In layoutlmv3_base_inference.yaml, the relevant dataset entry is publaynet/layout_scihub, did you train with that dataset?
Again, thanks for opening up your good source.
The text was updated successfully, but these errors were encountered:
hjbiao09
changed the title
About the
About Layout detection SFT dataset
Sep 12, 2024
Layoutlmv3 take COCO format data as input during the sft process in object detection. We use our own datasets for SFT, which would not be open-source. You can build your own sft dataset for your specific cases. Or you could provide those bad cases to us for further sft training. We will update our model once we collect enough bad cases.
Hi, first of all, thank you so much for releasing such a good source.
It works pretty well on our dataset, but we got incorrect layout detection in a few cases and want to further fine-tune the model with the layoutlmv3-sft pretrained model.
So, I would like to understand the format of the datasets used in the actual layout detection, could you please provide a link to the dataset source or related materials?
In
layoutlmv3_base_inference.yaml
, the relevant dataset entry ispublaynet/layout_scihub
, did you train with that dataset?Again, thanks for opening up your good source.
The text was updated successfully, but these errors were encountered: