Transformer-OCR icon indicating copy to clipboard operation
Transformer-OCR copied to clipboard

Results 15 Transformer-OCR issues
Sort by recently updated
recently updated
newest added

any information about training data structure?

Hi! I've download resnet101 from pytroch, as it was said in issue #1, but now i get this error: ![image](https://user-images.githubusercontent.com/44003762/100002229-7bf43380-2dee-11eb-8558-56d2872f2018.png) where can i get correct model for this project? Please,...

你好,我用自己的数据训练模型,如果图像resize成96*96大小,模型预测结果正常;如果resize成32×W大小,模型预测有问题,会识别成同一类,如下图: ![transform](https://user-images.githubusercontent.com/38058730/97281000-efe6fe80-1877-11eb-983d-1a12b21a9217.png) 请问这是什么原因呢?

when can we see the new paper ?

Is this a bug? ![image](https://user-images.githubusercontent.com/26408067/77746306-b4712100-7057-11ea-9745-0eb8f21f2159.png)

the src code assumes that the model is running the GPU, can I add a pull request for auto config?

can I make pull request for checking if the IMG value is None in __getitem__ function, and change image if True?

I applied the pre-trained model on ICDAR15 datasets, but the results are different from the reported ones in the paper?