… the potential of the Contrastive LanguageImage Pretraining (CLIP) model in scene text …
From two to one: A new scene text recognizer with visual language modeling network. In …

… and LM, we introduce a VisionLanguage Contrastive (VLC) module by encouraging positive
… of scene text recognition task. In this method, we propose a novel and powerful visual …

… text recognition in natural scenes (aka scene text recognition) … characteristics of the
corresponding languages. For example, … could learn a hybrid set of strokelets on multiple …

… of extracting visual features for scene text recognition. … language modeling module (IterLM)
in ABINet [10], our IterVM can iteratively enhance the visual features for scene text recognition…

Scene text recognition is a useful but very challenging task due to uncontrolled condition of
text in natural scenes. This paper presents a novel approach to recognize text in scene …

… review of works on scene text detection and recognition in the … ] in the fields of scene text
detection and recognition. However, … level (character detection) and high level (language prior) …

… the visual feature and the decoder focusing on the language … We propose SEED for scene
text recognition, which predicts … by the word embedding from a pre-trained language model. …

… as script identification, text/non-text classification and text-to-… directions on scene text
detection and recognition that need … Visual understanding: Vision and language is an interesting …

… , we also adopt a language model (LM) same as (Fang et al. 2021). Following the same
experimental setting, we first use ConCLR to pretrain a vision model (ABINet-Vision), and then …

… vision is suppressed so that no new information is acquired [37]. Inspired by these, we build
our scene text recognition … in recognition, we trained a 5-gram character language model (LM…