search

DevashishPrasad / CascadeTabNet

This repository contains the code and implementation details of the CascadeTabNet paper "CascadeTabNet: An approach for end to end table detection and structure recognition from image-based documents"
MIT License
1.46k stars 422 forks source link

demo error #112

Open Andrew05200 opened 3 years ago

Andrew05200 commented 3 years ago

image I run the demo on colab, config file is v2, but cannot get the results same as you. "show_result_pyplot" cannot get the input ('Bordered', 'cell', 'Borderless') in mmdetection2. can anyone give me some suggestions?

francescoperessini commented 3 years ago

@Andrew05200 I think the problem is due to the outdated version of the checkpoint. There is a tool to convert the checkpoint (https://github.com/open-mmlab/mmdetection/blob/master/tools/upgrade_model_version.py) to be correctly used with mmdetection2 but I was unable to use it. Let me know if you are able to do it!

jaytimbadia commented 3 years ago

image I run the demo on colab, config file is v2, but cannot get the results same as you. "show_result_pyplot" cannot get the input ('Bordered', 'cell', 'Borderless') in mmdetection2. can anyone give me some suggestions?

Hi, Hope you are doing great.

Can you please share your colab notebook since I am trying for 5 days and it's not even not working for me in colab. I am getting this error: ImportError: cannot import name 'isStringType'

Not sure how to resolve it, it would be really helpful if you can share your running notebook.

Thank you !!

francescoperessini commented 3 years ago

@jaytimbadia your code seems to be correct but, as the log says, the model and the loaded checkpoint don't match exactly. You have to convert the checkpoint to support MMDetectionV2+ and you can do that with the script in the folder Tools of this repository (https://github.com/DevashishPrasad/CascadeTabNet/tree/master/Tools).

Note that, as stated in the MMDetection doc, "The converted models can be run in MMDetection V2.0 with slightly dropped performance (less than 1% AP absolute)."

ghost commented 3 years ago

Hi @francescoperessini,

I was able to get rectangles on a right place after model update, however my annotations are still wrong. Any ideas, how to fix this? Thanks a lot for help. Capture

francescoperessini commented 3 years ago

Hi @francescoperessini,

I was able to get rectangles on a right place after model update, however my annotations are still wrong. Any ideas, how to fix this? Thanks a lot for help. Capture

Hi @Franky-Cranky,

you can easily change the classes' labels by modifying the parameter model.CLASSES, assigning it a new list of values; for example:

model.CLASSES = ['Bordered', 'cell', 'Borderless']

I don't remember the three labels' exact order; you have to try!