Token label matching in layoutlmv3 inference

[vidya-chandran]

@NielsRogge Hi can you help me on how to find a mapping between the predicted labels and tokens(words). I have got true boxes and true predictions. I also need which all words belong to each of the labels. I tried extracting the words within bounding boxes from the inference image, but it gave incorrect words (may be problem with the technique I used). Please give an explaination on this. Many thanks in advance