SCATTER: Selective Context Attentional Scene Text Recognizer

Abstract

Scene Text Recognition (STR), the task of recognizing text against complex image backgrounds, is an active area of research. Current state-of-the-art (SOTA) methods still struggle to recognize text written in arbitrary shapes. In this paper, we introduce a novel architecture for STR, named Selective Context ATtentional Text Recognizer (SCATTER). SCATTER utilizes a stacked block architecture with intermediate supervision during training, that paves the way to successfully train a deep BiLSTM encoder, thus improving the encoding of contextual dependencies. Decoding is done using a two-step 1D attention mechanism. The first attention step re-weights visual features from a CNN backbone together with contextual features computed by a BiLSTM layer. The second attention step, similar to previous papers, treats the features as a sequence and attends to the intra-sequence relationships. Experiments show that the proposed approach surpasses SOTA performance on irregular text recognition benchmarks by 3.7% on average.

Sequence-to-Sequence Contrastive Learning for Text Recognition

“…Un-and semi-supervised learning for text recognition Despite clear advantages, currently, most text recognition methods do not utilize unlabeled real-world text images. Specifically, handwritten recognition usually relies on fullysupervised training [64,59], while scene text models are trained mostly on synthetic data [3,38]. That said, [68] and [34] have recently suggested domain adaptation techniques to utilize an unlabeled dataset along with labeled data.…”

Random Blur Data Augmentation for Scene Text Recognition

“…Similar to [18], many works [19]- [21] also introduce BiLSTM into the STR framework. Litman et al [22] propose to apply a deeper BiLSTM model to improve the encoding of contextual and deploy intermediate supervisions along the network layers. Different from [22], some works [23]- [25] do not use BiLSTM structure, and consider that BiLSTM is computationally intensive and time consuming.…”

“…Litman et al [22] propose to apply a deeper BiLSTM model to improve the encoding of contextual and deploy intermediate supervisions along the network layers. Different from [22], some works [23]- [25] do not use BiLSTM structure, and consider that BiLSTM is computationally intensive and time consuming. Yin et al [23] propose to simultaneously detect and recognize characters by sliding the text line image with character models, which are learned end-to-end on text line images labeled with text transcripts.…”

NASTER: Non-local Attentional Scene Text Recognizer

“…Zhan et al [39] improved the recognition performance by iteratively correcting the text image. Litman et al [17] improved the performance of recognition by overlaying the RNN decoder several times. They demonstrated the effectiveness of the superimposed module.…”