JiaquanYe / TableMASTER-mmocr

2nd solution of ICDAR 2021 Competition on Scientific Literature Parsing, Task B.
Apache License 2.0
442 stars 104 forks source link

Structure recognition accuracy #34

Open cqray1990 opened 3 years ago

cqray1990 commented 3 years ago

i want to only test the acc of Structure recognition accuracy and use mmocr_teds_acc_mp.py, then i use pretrain model name as you supply epoch_16_0.7767.pth, to get only table structure results when run mmocr_teds_acc_mp.py the AVG TEDS score is 0.327583870638436 not 0.78,

TableMASTER_maxlength_500 | 0.7808

JiaquanYe commented 3 years ago

i want to only test the acc of Structure recognition accuracy and use mmocr_teds_acc_mp.py, then i use pretrain model name as you supply epoch_16_0.7767.pth, to get only table structure results when run mmocr_teds_acc_mp.py the AVG TEDS score is 0.327583870638436 not 0.78,

TableMASTER_maxlength_500 | 0.7808

TEDS score measures both the structure similarity and the cell content similarity between the prediction and the ground truth. And 0.7808 is the word accuracy of TableMASTER. If you want to get TEDS score, you must get PSENet detection results, TableMASTER structure recognition results and MASTER textline recognition results, and use match.py to get TEDS score.

tangjiashan7 commented 3 years ago

i want to only test the acc of Structure recognition accuracy and use mmocr_teds_acc_mp.py, then i use pretrain model name as you supply epoch_16_0.7767.pth, to get only table structure results when run mmocr_teds_acc_mp.py the AVG TEDS score is 0.327583870638436 not 0.78,

TableMASTER_maxlength_500 | 0.7808

The score of the test structure alone is the same as said. But I have a question for you how do you test the structure of a single image