Focusing Attention: Towards Accurate Text Recognition in Natural Images
arXiv:1709.02054 · doi:10.1109/ICCV.2017.543
Abstract
Scene text recognition has been a hot research topic in computer vision due to its various applications. The state of the art is the attention-based encoder-decoder framework that learns the mapping between input images and output sequences in a purely data-driven way. However, we observe that existing attention-based methods perform poorly on complicated and/or low-quality images. One major reason is that existing methods cannot get accurate alignments between feature areas and targets for such images. We call this phenomenon "attention drift". To tackle this problem, in this paper we propose the FAN (the abbreviation of Focusing Attention Network) method that employs a focusing attention mechanism to automatically draw back the drifted attention. FAN consists of two major components: an attention network (AN) that is responsible for recognizing character targets as in the existing methods, and a focusing network (FN) that is responsible for adjusting attention by evaluating whether AN pays attention properly on the target areas in the images. Furthermore, different from the existing methods, we adopt a ResNet-based network to enrich deep representations of scene text images. Extensive experiments on various benchmarks, including the IIIT5k, SVT and ICDAR datasets, show that the FAN method substantially outperforms the existing methods.
arXiv commentRevise the description of IC15 datasets (1811 samples)
References in corpus (8)
- Neural Machine Translation by Jointly Learning to Align and Translate
- Sequence to Sequence Learning with Neural Networks
- ADADELTA: An Adaptive Learning Rate Method
- Caffe: Convolutional Architecture for Fast Feature Embedding
- Attention-Based Models for Speech Recognition
- Synthetic Data and Artificial Neural Networks for Natural Scene Text Recognition
- End-to-End Text Recognition with Hybrid HMM Maxout Models
- Joint CTC-Attention based End-to-End Speech Recognition using Multi-task Learning
Cited by in corpus (99)
- MASTER: Multi-Aspect Non-local Network for Scene Text Recognition
- Scene Text Detection and Recognition: The Deep Learning Era
- What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis
- Convolutional Neural Networks with Gated Recurrent Connections
- Recognition of Handwritten Chinese Text by Segmentation: A Segment-annotation-free Approach
- Towards Accurate Scene Text Recognition with Semantic Reasoning Networks
- CLIP4STR: A Simple Baseline for Scene Text Recognition with Pre-trained Vision-Language Model
- 2D Attentional Irregular Scene Text Recognizer
- Text Recognition in the Wild: A Survey
- TextSR: Content-Aware Text Super-Resolution Guided by Recognition
- Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes
- Show, Attend and Read: A Simple and Strong Baseline for Irregular Text Recognition
- Scene Text Recognition from Two-Dimensional Perspective
- Scene Text Synthesis for Efficient and Effective Deep Network Training
- ESIR: End-to-end Scene Text Recognition via Iterative Image Rectification
- 2D-CTC for Scene Text Recognition
- Text Detection and Recognition in the Wild: A Review
- FACLSTM: ConvLSTM with Focused Attention for Scene Text Recognition
- Towards Unconstrained End-to-End Text Spotting
- Read Like Humans: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Recognition
- SEED: Semantics Enhanced Encoder-Decoder Framework for Scene Text Recognition
- NRTR: A No-Recurrence Sequence-to-Sequence Model For Scene Text Recognition
- CDistNet: Perceiving Multi-Domain Character Distance for Robust Text Recognition
- Word-level Sign Language Recognition with Multi-stream Neural Networks Focusing on Local Regions and Skeletal Information
- Focus-Enhanced Scene Text Recognition with Deformable Convolutions
- RobustScanner: Dynamically Enhancing Positional Clues for Robust Text Recognition
- TRIE: End-to-End Text Reading and Information Extraction for Document Understanding
- TextScanner: Reading Characters in Order for Robust Scene Text Recognition
- A Multi-Object Rectified Attention Network for Scene Text Recognition
- TRIG: Transformer-Based Text Recognizer with Initial Embedding Guidance
- Recurrent Calibration Network for Irregular Text Recognition
- Offline Handwritten Chinese Text Recognition with Convolutional Neural Networks
- From Two to One: A New Scene Text Recognizer with Visual Language Modeling Network
- Rethinking Irregular Scene Text Recognition
- Convolutional Character Networks
- Generating Text Sequence Images for Recognition
- Double Supervised Network with Attention Mechanism for Scene Text Recognition
- Handwritten Mathematical Expression Recognition with Bidirectionally Trained Transformer
- Bidirectional Scene Text Recognition with a Single Decoder
- Learn to Augment: Joint Data Augmentation and Network Optimization for Text Recognition
- Decoupled Attention Network for Text Recognition
- SCATTER: Selective Context Attentional Scene Text Recognizer
- Improving Attention-Based Handwritten Mathematical Expression Recognition with Scale Augmentation and Drop Attention
- Self-Supervised Learning for Text Recognition: A Critical Survey
- Towards End-to-End Text Spotting in Natural Scenes
- Gaussian Constrained Attention Network for Scene Text Recognition
- Scene Text Image Super-Resolution in the Wild
- Chinese Street View Text: Large-scale Chinese Text Reading with Partially Supervised Learning
- You Only Recognize Once: Towards Fast Video Text Spotting
- Hamming OCR: A Locality Sensitive Hashing Neural Network for Scene Text Recognition
- PERT: A Progressively Region-based Network for Scene Text Removal
- GA-DAN: Geometry-Aware Domain Adaptation Network for Scene Text Detection and Recognition
- SAFL: A Self-Attention Scene Text Recognizer with Focal Loss
- STRIDE : Scene Text Recognition In-Device
- Primitive Representation Learning for Scene Text Recognition
- Revisiting Classification Perspective on Scene Text Recognition
- A Holistic Representation Guided Attention Network for Scene Text Recognition
- AutoSTR: Efficient Backbone Search for Scene Text Recognition
- Text Perceptron: Towards End-to-End Arbitrary-Shaped Text Spotting
- A Hybrid Vision Transformer Approach for Mathematical Expression Recognition
- Separating Content from Style Using Adversarial Learning for Recognizing Text in the Wild
- Joint Visual Semantic Reasoning: Multi-Stage Decoder for Text Recognition
- Spatial Context-based Self-Supervised Learning for Handwritten Text Recognition
- MetaHTR: Towards Writer-Adaptive Handwritten Text Recognition
- SPIN: Structure-Preserving Inner Offset Network for Scene Text Recognition
- GTC: Guided Training of CTC Towards Efficient and Accurate Scene Text Recognition
- Refined Gate: A Simple and Effective Gating Mechanism for Recurrent Units
- SNIDER: Single Noisy Image Denoising and Rectification for Improving License Plate Recognition
- Character Region Attention For Text Spotting
- AON: Towards Arbitrarily-Oriented Text Recognition
- Towards the Unseen: Iterative Text Recognition by Distilling from Errors
- I2C2W: Image-to-Character-to-Word Transformers for Accurate Scene Text Recognition
- Text is Text, No Matter What: Unifying Text Recognition using Knowledge Distillation
- Parallel Scale-wise Attention Network for Effective Scene Text Recognition
- On Vocabulary Reliance in Scene Text Recognition
- Simultaneous Recognition of Horizontal and Vertical Text in Natural Images
- Decoupling Visual-Semantic Feature Learning for Robust Scene Text Recognition
- Representation and Correlation Enhanced Encoder-Decoder Framework for Scene Text Recognition
- Towards Fully Automated Manga Translation
- AE TextSpotter: Learning Visual and Linguistic Representation for Ambiguous Text Spotting
- ReADS: A Rectified Attentional Double Supervised Network for Scene Text Recognition
- Improving Structured Text Recognition with Regular Expression Biasing
- Object-QA: Towards High Reliable Object Quality Assessment
- Towards Real-Time Advancement of Underwater Visual Quality with GAN
- PIMNet: A Parallel, Iterative and Mimicking Network for Scene Text Recognition
- Adaptive Embedding Gate for Attention-Based Scene Text Recognition
- IFR: Iterative Fusion Based Recognizer For Low Quality Scene Text Recognition
- Text Recognition in Real Scenarios with a Few Labeled Samples
- Using Human Psychophysics to Evaluate Generalization in Scene Text Recognition Models
- SAFE: Scale Aware Feature Encoder for Scene Text Recognition
- Scene Text recognition with Full Normalization
- Exploring Font-independent Features for Scene Text Recognition
- Reciprocal Feature Learning via Explicit and Implicit Tasks in Scene Text Recognition
- Towards Pure End-to-End Learning for Recognizing Multiple Text Sequences from an Image
- Boosting Scene Character Recognition by Learning Canonical Forms of Glyphs
- Scene Text Recognition With Finer Grid Rectification
- Stratified Domain Adaptation: A Progressive Self-Training Approach for Scene Text Recognition
- Scene Text Recognition with Temporal Convolutional Encoder
- Context-Free TextSpotter for Real-Time and Mobile End-to-End Text Detection and Recognition