How to annotate the own receipt images for layoutLM

[SandyRSK]

I have download the SROIE dataset. It has Box, img, key folder Once the pre-processing is done.

I run !python "run_seq_labeling.py" --data_dir "output_receipt" --model_type layoutlm --model_name_or_path "model_LM/" --do_lower_case --max_seq_length 512 --do_train --num_train_epochs 2 --logging_steps 10 --save_steps -1 --output_dir "out_model_receipt/" --overwrite_output_dir --labels "labels_receipt.txt" --per_gpu_train_batch_size 16 --per_gpu_eval_batch_size 16 --fp16

Error occurred is :

File "/content/drive/My Drive/unilm/layoutlm/examples/seq_labeling/run_seq_labeling.py", line 811, in main() File "/content/drive/My Drive/unilm/layoutlm/examples/seq_labeling/run_seq_labeling.py", line 701, in main args, tokenizer, labels, pad_token_label_id, mode="train" File "/content/drive/My Drive/unilm/layoutlm/examples/seq_labeling/layoutlm/data/funsd.py", line 72, in init self.all_bboxes = torch.tensor([f.boxes for f in features], dtype=torch.long) ValueError: expected sequence of length 4 at dim 2 (got 8)

[wolfshow]

@SandyRSK Basically, you may need pre-process the SROIE dataset into token-level and fed the data into LayoutLM.

[SandyRSK]

@wolfshow

After all the preprocessing of my own receipt dataset, I can able to train. But in the prediction stage, after running the --do predict program I can able to see the text_prediction.txt file but it shows only top 16. I don't know Why it is not detecting the other?

[ruifcruz]

Mabe it helps: https://github.com/ruifcruz/sroie-on-layoutlm

[burhanuddin03]

I have some invoice dataset that I want to annotate in order to run it through layoutlm but the problem is where should I annotate it, I couldn't find a tool which takes in a image of document allow me to annotate it and return me the text files that I can further feed into layoutlm. I tried UBIAI but that tool is paid. Could anyone suggest me something similar to ubiai?

[Irisnotiris]

I have some invoice dataset that I want to annotate in order to run it through layoutlm but the problem is where should I annotate it, I couldn't find a tool which takes in a image of document allow me to annotate it and return me the text files that I can further feed into layoutlm. I tried UBIAI but that tool is paid. Could anyone suggest me something similar to ubiai?

Hi, any progress on annotation tools? also tried but coundn't find others except UBIAI...

[Rajeshwar21]

I am

I have some invoice dataset that I want to annotate in order to run it through layoutlm but the problem is where should I annotate it, I couldn't find a tool which takes in a image of document allow me to annotate it and return me the text files that I can further feed into layoutlm. I tried UBIAI but that tool is paid. Could anyone suggest me something similar to ubiai?

Hi, any progress on annotation tools? also tried but coundn't find others except UBIAI...

Have u found any such labeling tool? Even I'm looking for the same

[burhanuddin03]

I am

I have some invoice dataset that I want to annotate in order to run it through layoutlm but the problem is where should I annotate it, I couldn't find a tool which takes in a image of document allow me to annotate it and return me the text files that I can further feed into layoutlm. I tried UBIAI but that tool is paid. Could anyone suggest me something similar to ubiai?

Hi, any progress on annotation tools? also tried but coundn't find others except UBIAI...

Have u found any such labeling tool? Even I'm looking for the same

No, there is a option on UBIAI to negotiate on the budget. I used that .

[ammarsaf]

I have some invoice dataset that I want to annotate in order to run it through layoutlm but the problem is where should I annotate it, I couldn't find a tool which takes in a image of document allow me to annotate it and return me the text files that I can further feed into layoutlm. I tried UBIAI but that tool is paid. Could anyone suggest me something similar to ubiai?

I am not sure it is full related or not but I found Layout Parser and Label Studio.

Link: https://www.youtube.com/watch?v=puOKTFXRyr4&ab_channel=LabelStudio