search

microsoft / TAP

TAP: Text-Aware Pre-training for Text-VQA and Text-Caption, CVPR 2021 (Oral)
MIT License
70 stars 11 forks source link

About the number of OCR in stvqa dataset #4

Closed JayZhu0104 closed 2 years ago

JayZhu0104 commented 2 years ago

Hi！ I found that the number of words detected by OCR in some pictures in stvqa dataset is inconsistent with the corresponding feature number. For example, the number of features in 'featresx/stvqa/train/imageNet/n03196217 7957. npy' is 33, while the number of OCR words in the corresponding 'ocr feat resx/stvqa conf/train/imageNet/n03196217 7957_info. npy' is 55. The two numbers do not match. About 2000 pictures have this problem in train dataset. image

zyang-ur commented 2 years ago

Updated the corresponded files :)