DAVAR-Lab-OCR

DAVAR-Lab-OCR copied to clipboard

TRIE test error, not use GT

4

Thank you for releasing the nice code! I try e2e_trie but not use GT_bboxes for inference and have this problem when computing the f1-score  I think the number of...

ZHEGG

RF-Learning中的visual和total指标问题

1

再https://github.com/hikopensource/DAVAR-Lab-OCR/issues/65 中，提到RF-visual仅训练了计数分支，那在readme中的指标从何而来

WenmuZhou

def html_to_area(html_str, row_index, span_matrix): """Convert html to span matrix, a two-dimensional matrix representing table structure Args: html_str(str): html representing table structure. row_index(list): index of each row in html. span_matrix(np.array): a...

cqray1990

cqray1990

cqray1990

Arlen-yuzu

how to convert LGPMA to onnx

2

cqray1990

How large are the training parameters of the VSR model? At present, I only have two video cards with 8G video Memory, which always show out of Memory. How can...

369-SN

您好，关于为LGPMA构建训练数据，我有以下两点疑问： 1、是否一定需要包含空单元格（blank cell）的标注； 2、bbox是否一定要按照pubtabnet那样标注为text region的格式，如果标注为align cell bound，会对模型产生怎样的负面影响（因为lgpma的lpma分支里面有对text region的mask predict） 可否解答我的疑惑，谢谢

Arlen-yuzu

The code opensourced in this repo seems not open the refine_bboxes during post-processing.

Arlen-yuzu