Saved in:
Bibliographic Details
Main Authors: Qiao, Yixuan, Chen, Hao, Wang, Jun, Zhao, Shanshan, Chen, Yihao, Ye, Xianbin, Li, Ziliang, Qi, Xianbiao, Gao, Peng, Xie, Guotong
Format: Preprint
Published: 2021
Subjects:
Online Access:https://arxiv.org/abs/2106.15332
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915785200369664
author Qiao, Yixuan
Chen, Hao
Wang, Jun
Zhao, Shanshan
Chen, Yihao
Ye, Xianbin
Li, Ziliang
Qi, Xianbiao
Gao, Peng
Xie, Guotong
author_facet Qiao, Yixuan
Chen, Hao
Wang, Jun
Zhao, Shanshan
Chen, Yihao
Ye, Xianbin
Li, Ziliang
Qi, Xianbiao
Gao, Peng
Xie, Guotong
contents TextVQA requires models to read and reason about text in images to answer questions about them. Specifically, models need to incorporate a new modality of text present in the images and reason over it to answer TextVQA questions. In this challenge, we use generative model T5 for TextVQA task. Based on pre-trained checkpoint T5-3B from HuggingFace repository, two other pre-training tasks including masked language modeling(MLM) and relative position prediction(RPP) are designed to better align object feature and scene text. In the stage of pre-training, encoder is dedicate to handle the fusion among multiple modalities: question text, object text labels, scene text labels, object visual features, scene visual features. After that decoder generates the text sequence step-by-step, cross entropy loss is required by default. We use a large-scale scene text dataset in pre-training and then fine-tune the T5-3B with the TextVQA dataset only.
format Preprint
id arxiv_https___arxiv_org_abs_2106_15332
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle Winner Team Mia at TextVQA Challenge 2021: Vision-and-Language Representation Learning with Pre-trained Sequence-to-Sequence Model
Qiao, Yixuan
Chen, Hao
Wang, Jun
Zhao, Shanshan
Chen, Yihao
Ye, Xianbin
Li, Ziliang
Qi, Xianbiao
Gao, Peng
Xie, Guotong
Computer Vision and Pattern Recognition
TextVQA requires models to read and reason about text in images to answer questions about them. Specifically, models need to incorporate a new modality of text present in the images and reason over it to answer TextVQA questions. In this challenge, we use generative model T5 for TextVQA task. Based on pre-trained checkpoint T5-3B from HuggingFace repository, two other pre-training tasks including masked language modeling(MLM) and relative position prediction(RPP) are designed to better align object feature and scene text. In the stage of pre-training, encoder is dedicate to handle the fusion among multiple modalities: question text, object text labels, scene text labels, object visual features, scene visual features. After that decoder generates the text sequence step-by-step, cross entropy loss is required by default. We use a large-scale scene text dataset in pre-training and then fine-tune the T5-3B with the TextVQA dataset only.
title Winner Team Mia at TextVQA Challenge 2021: Vision-and-Language Representation Learning with Pre-trained Sequence-to-Sequence Model
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2106.15332