Skip to content

Select language

目前語言: 繁體中文

Deep Text Detection and Recognition

Source: https://github.com/hwalsuklee/awesome-deep-text-detection-recognition at 0958b3875043
License: Apache-2.0
Read the English original

This edition is an AI-assisted translation.

精選深度學習文字偵測與辨識論文清單。

  • 論文依發表日期排序。
  • IC 是 ICDAR 的縮寫。
  • 分數為定位任務的 F1 分數。
    • (L) 表示排行榜上的分數。
    • 若排行榜公布的分數與論文略有不同,則提供 (L) 分數。
  • *CODE 表示官方程式碼,CODE(M) 表示提供訓練完成的模型。
會議日期標題IC13IC15資源
‘14-ECCV14/10/07Robust Scene Text Detection with Convolution Neural Network Induced MSER Trees
15-CVPR15/06/01Symmetry-based text line detection in natural scenes0.8043PRJ
CODE
‘16-TIP15/10/12Text-Attentional Convolutional Neural Networks for Scene Text Detection0.8165
‘15-ICCV15/12/13Text Flow : A Unified Text Detection System in Natural Scene Images0.8025
‘16-arXiv16/03/31Accurate Text Localization in Natural Image with Cascaded Convolutional TextNetwork0.86
‘16-CVPR16/04/14Multi-Oriented Text Detection with Fully Convolutional Networks0.830.54*TORCH(M)
‘16-CVPR16/04/22Synthetic Data for Text Localisation in Natural Images0.847
(L)0.8359
CODE
DB
‘16-arXiv16/06/29Scene Text Detection Via Holistic, Multi-Channel Prediction0.84330.6477
‘16-ECCV16/09/12Detecting Text in Natural Image with Connectionist Text Proposal Network0.82150.6085*CAFFE(M)
CAFFE
TF(M)
TF
DEMO
BLOG(CH)
‘17-AAAI16/11/21TextBoxes: A fast text detector with a single deep neural network0.85
(L)0.8767
*CAFFE(M)
TF
BLOG(KR)
‘18-TM17/03/03Arbitrary-Oriented Scene Text Detection via Rotation Proposals0.91250.8020*CAFFE
‘17-CVPR17/03/04Deep Matching Prior Network: Toward Tighter Multi-oriented Text Detection0.7064
‘17-CVPR17/03/19Detecting Oriented Text in Natural Images by Linking Segments0.8530.75
(L)0.7636
*TF(M)
TF(M)
SLIDE
VIDEO
‘17-arXiv17/03/24Deep Direct Regression for Multi-Oriented Scene Text Detection0.860.81
‘17-arXiv17/04/03Cascaded Segmentation-Detection Networks for Word-Level Text Spotting0.860.71
‘17-CVPR17/04/11EAST: An Efficient and Accurate Scene Text Detector0.8072
(L)0.8038
TF(M)
TF
PYTORCH(M)
PYTORCH
DEMO
KERAS(M)
VIDEO
‘17-ICIP17/05/15WordFence: Text Detection in Natural Images with Border Awareness0.86
‘17-arXiv17/06/30R2CNN: Rotational Region CNN for Orientation Robust Scene Text Detection0.87730.8254TF(M)
CAFFE(M)
‘17-CVPR17/07/21Multi-scale FCN with Cascaded Instance Aware Segmentation for Arbitrary Oriented Word Spotting In The Wild0.850.63
‘17-arXiv17/08/17Deep Scene Text Detection with Connected Component Proposals0.919
‘17-ICCV17/08/22WordSup: Exploiting Word Annotations for Character based Text Detection0.90640.7816
‘17-ICCV17/09/01Single Shot Text Detector with Regional Attention0.87040.7691*CAFFE(M)
PYTORCH
VIDEO
‘17-arXiv17/09/11Fused Text Segmentation Networks for Multi-oriented Scene Text Detection0.8414
‘17-ICCV17/10/13WeText: Scene Text Detection under Weak Supervision0.869
(L)0.8313
‘17-ICCV17/10/22Self-organized Text Detection with Minimal Post-processing via Border Learning0.84*KERAS(M)
‘17-ICDAR17/11/11Deep Residual Text Detection Network for Scene Text0.9117
(L)0.8925
‘18-AAAI17/11/12Feature Enhancement Network: A Refined Scene Text Detector0.9161
‘17-arXiv17/11/30ArbiText: Arbitrary-Oriented Text Detection in Unconstrained Scene0.759
‘18-AAAI18/01/04PixelLink: Detecting Scene Text via Instance Segmentation0.8810.8519*TF(M) TF
‘18-CVPR18/01/05FOTS: Fast Oriented Text Spotting with a Unified Network0.9250.8984PYTORCH
PYTORCH
VIDEO
‘18-TIP18/01/09TextBoxes++: A Single-Shot Oriented Scene Text Detector0.880.829
(L)0.8475
*CAFFE(M)
‘18-CVPR18/02/27Multi-Oriented Scene Text Detection via Corner Localization and Region Segmentation0.880.843*PYTORCH(M)
‘18-CVPR18/03/09An end-to-end TextSpotter with Explicit Alighment and Attention0.90.87*CAFFE(M)
‘18-CVPR18/03/14Rotation-Sensitive Regression for Oriented Scene Text Detection0.890.838*CAFFE(M)
‘18-arXiv18/04/08Detecting Multi-Oriented Text with Corner-based Region Proposals0.8760.845*CAFFE(M)
‘18-arXiv18/04/24An Anchor-Free Region Proposal Network for Faster R-CNN based Text Detection Approaches0.920.86
‘18-IJCAI18/05/03IncepText: A New Inception-Text Module with Deformable PSROI Pooling for Multi-Oriented Scene Text Detection0.9047
‘18-arXiv18/06/07Shape Robust Text Detection with Progressive Scale Expansion Network0.8721PRJ
‘18-ECCV18/07/04TextSnake: A Flexible Representation for Detecting Text of Arbitrary Shapes0.826PYTORCH
‘18-ECCV18/07/06Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes0.9170.86
‘18-ECCV18/07/10Accurate Scene Text Detection through Border Semantics Awareness and Bootstrapping0.892
‘19-AAAI18/11/21Scene Text Detection with Supervised Pyramid Context Network0.9210.872
‘19-TIP18/12/04TextField: Learning A Deep Direction Field for Irregular Scene Text Detection0.824*CAFFE(M)
‘19-CVPR19/03/21Towards Robust Curve Text Detection with Conditional Spatial Expansion
‘19-CVPR19/03/28Shape Robust Text Detection with Progressive Scale Expansion Network0.857TF(M)
‘19-CVPR19/04/03Character Region Awareness for Text Detection0.9520.869*PYTORCH(M)
VIDEO
PYTORCH
TF(M)
KERAS
BLOG_CH
BLOG_KR
BLOG_KR
BLOG_KR
‘19-CVPR19/04/13Look More Than Once: An Accurate Detector for Text of Arbitrary Shapes Screen reader support enabled0.877
‘19-CVPR19/06/16Learning Shape-Aware Embedding for Scene Text Detection0.877
‘19-CVPR19/06/16Arbitrary Shape Scene Text Detection with Adaptive Text Region Representation0.9170.876
‘19-ICCV19/08/16Efficient and Accurate Arbitrary-Shaped Text Detection with Pixel Aggregation Network0.829
‘19-ICCV19/09/02Geometry Normalization Networks for Accurate Scene Text Detection0.8852
‘19-AAAI19/11/20Real-time Scene Text Detection with Differentiable Binarization0.847

  • 論文依發表日期排序。
  • IC 是 ICDAR 的縮寫。
  • 分數為辨識任務的詞彙準確率。
    • IC03、IC13 與 IC15 資料集的結果中,各論文使用的樣本數不同,
      但我們未加以區分。
  • *CODE 表示官方程式碼,CODE(M) 表示提供訓練完成的模型。
會議日期標題SVTIIIT5kIC03IC13資源
‘15-ICLR14/12/18Deep structured output learning for unconstrained text recognition0.7170.8960.818TF
SLIDE
VIDEO
‘16-IJCV15/05/07Reading text in the wild with convolutional neural networks0.8070.9330.908KERAS
‘16-AAAI15/06/14Reading Scene Text in Deep Convolutional Sequences
‘17-TPAMI15/07/21An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition0.8080.7820.8940.867TORCH(M)
TF
TF
TF
TF
PYTORCH
PYTORCH(M)
BLOG(KR)
‘16-CVPR16/03/09Recursive Recurrent Nets with Attention Modeling for OCR in the Wild0.8070.7840.8870.9
‘16-CVPR16/03/12Robust scene text recognition with automatic rectification0.8190.8190.9010.886PYTORCH
PYTORCH
‘16-CVPR16/06/27CNN-N-Gram for Handwriting Word Recognition0.8362VIDEO
‘16-BMVC16/09/19STAR-Net: A SpaTial Attention Residue Network for Scene Text Recognition0.8360.8330.8990.891
‘17-arXiv17/07/27STN-OCR: A single Neural Network for Text Detection and Text Recognition0.7980.860.903*MXNET(M)
PRJ
BLOG
‘17-IJCAI17/08/19Learning to Read Irregular Text with Attention Mechanisms
‘17-arXiv17/09/06Scene Text Recognition with Sliding Convolutional Character Models0.7650.8160.8450.852
‘17-ICCV17/09/07Focusing Attention: Towards Accurate Text Recognition in Natural Images0.8590.8740.9420.933
‘18-CVPR17/11/12AON: Towards Arbitrarily-Oriented Text Recognition0.8280.870.915TF
‘17-NIPS17/12/04Gated Recurrent Convolution Neural Network for OCR0.8150.8080.978*TORCH(M)
‘18-AAAI18/01/04Char-Net: A Character-Aware Neural Network for Distorted Scene Text Recognition0.8440.8360.9150.908
‘18-AAAI18/01/04SqueezedText: A Real-time Scene Text Recognition by Binary Convolutional Encoder-decoder Network0.870.9310.929
‘18-CVPR18/05/09Edit Probability for Scene Text Recognition0.8750.8830.9460.944
‘18-TPAMI18/06/25ASTER: An Attentional Scene Text Recognizer with Flexible Rectification0.9360.9340.9450.918*TF(M)
PYTORCH
‘18-ECCV18/09/08Synthetically Supervised Feature Learning for Scene Text Recognition0.8710.8940.9470.94
‘19-AAAI18/09/18Scene Text Recognition from Two-Dimensional Perspective0.8210.920.914
‘19-AAAI18/11/02Show, Attend and Read: A Simple and Strong Baseline for Irregular Text Recognition0.8450.9150.91*TORCH(M)
‘19-CVPR18/12/14ESIR: End-to-end Scene Text Recognition via Iterative Image Rectification0.9020.9330.913PRJ
‘19-PR19/01/10MORAN: A Multi-Object Rectified Attention Network for Scene Text Recognition0.8830.9120.9500.924*PYTORCH(M)
‘19-ICCV19/04/03What is wrong with scene text recognition model comparisons? dataset and model analysis0.8750.9490.936*PYTORCH(M)
BLOG_KR
‘19-CVPR19/04/18Aggregation Cross-Entropy for Sequence Recognition0.8260.8230.9210.897*PYTORCH
‘19-CVPR19/06/16Sequence-to-Sequence Domain Adaptation Network for Robust Text Image Recognition0.8450.8380.9210.918
‘19-ICCV19/08/06Symmetry-constrained Rectification Network for Scene Text Recognition0.8890.9440.950.939
‘20-AAAI19/12/28TextScanner: Reading Characters in Order for Robust Scene Text Recognition0.8950.9260.925
‘20-AAAI19/12/21Decoupled Attention Network for Text Recognition0.8920.9430.950.939*PYTORCH(M)
‘20-AAAI20/02/04GTC: Guided Training of CTC0.9290.9550.9520.943

  • 論文依發表日期排序。
  • IC 是 ICDAR 的縮寫。
  • 分數為通用任務的 F1 分數。
  • *CODE 表示官方程式碼,CODE(M) 表示提供訓練完成的模型。
會議日期標題IC03IC13IC15資源
‘12-ICPR12/11/11End-to-end text recognition with convolutional neural networks0.67*CODE
‘14-ECCV14/09/06Deep Features for Text Spotting0.75PRJ
MATLAB
‘15-IJCV15/05/07Reading Text in the Wild with Convolutional Neural Networks0.700.77KERAS
‘15-TPAMI15/10/30Real-time Lexicon-free Scene Text Localization and Recognition0.5420.156
‘16-arXiv16/04/10TextProposals: a Text-specific Selective Search Algorithm for Word Spotting in the Wild0.68430.4718
(L)0.533
*CAFFE(M)
‘17-AAAI16/11/21TextBoxes: A fast text detector with a single deep neural network0.84TF
*CAFFE(M)
BLOG_KR
‘17-ICCV17/07/13Towards End-to-end Text Spotting with Convolution Recurrent Neural Network0.8459VIDEO
‘17-ICCV17/10/22Deep TextSpotter An End-to-End Trainable Scene Text Localization and Recognition Framework0.770.47VIDEO
*CAFFE(M)
‘18-CVPR18/01/05FOTS: Fast Oriented Text Spotting with a Unified Network0.84770.6533VIDEO
TF(M)
‘18-TIP18/01/09TextBoxes++: A Single-Shot Oriented Scene Text Detector0.84650.519*CAFFE(M)
‘18-CVPR18/03/09An end-to-end TextSpotter with Explicit Alignment and Attention0.860.63*CAFFE(M)
‘18-TPAMI18/06/25ASTER: An Attentional Scene Text Recognizer with Flexible Rectification0.64*TF(M)
‘18-ECCV18/07/06Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes0.8650.624
‘19-ICCV19/08/24Towards Unconstrained End-to-End Text Spotting0.6994BLOG_KR
‘19-ICCV19/10/17Convolutional Character Networks0.7108*PYTORCH(M)
‘19-ICCV19/10/27TextDragon: An End-to-End Framework for Arbitrary Shaped Text Spotting0.6537
‘20-AAAI19/11/21All You Need Is Boundary: Toward Arbitrary-Shaped Text Spotting0.8410.641
‘20-AAAI20/02/12Text Perceptron: Towards End-to-End Arbitrary-Shaped Text Spotting0.8580.651

  • 論文依發表日期排序。
  • *CODE 表示官方程式碼,CODE(M) 表示提供訓練完成的模型。
會議日期標題說明資源
‘14-NIPS14/06/09Synthetic Data and Artificial Neural Networks for Natural Scene Text Recognition資料集PRJ
‘17-ECCV17/02/13End-to-End Interpretation of the French Street Name Signs Dataset資料集(FSNS)*TF(M)
‘17-arXiv17/04/11Attention-based Extraction of Structured Information from Street View ImageryFSNS*TF(M)
TF
TF
LUA
BLOG_KR
‘17-CVPR17/07/21Unambiguous Text Localization and Retrieval for Cluttered Scenes文字檢索
‘17-AAAI17/10/22Detection and Recognition of Text Embedded in Online Images via Neural Context Models資料集PRJ
‘18-CVPR17/11/17Separating Style and Content for Generalized Style Transfer字型樣式
‘17-arXiv17/12/06Detecting Curve Text in the Wild New Dataset and New Solution資料集(CTW 1500)PRJ
‘18-AAAI17/12/14SEE: Towards Semi-Supervised End-to-End Scene Text RecognitionFSNSPRJ
*CHAINER(M)
‘17-CVPR18/06/07Learning to Extract Semantic Structure from Documents Using Multimodal Fully Convolutional Neural Networks文件版面PRJ
‘18-CVPR18/06/19DocUNet: Document Image Unwarping via A Stacked U-Net文件去扭曲PRJ
‘18-CVPR18/06/19Document Enhancement using Visibility Detection文件增強PRJ
‘18-IJCAI18/06/22Multi-Task Handwritten Document Layout Analysis文件版面
‘18-ECCV18/07/09Verisimilar Image Synthesis for Accurate Detection and Recognition of Texts in Scenes資料集PRJ
‘19-AAAI18/12/03EnsNet: Ensconce Text in the Wild文字移除DB
‘19-CVPR18/12/14Spatial Fusion GAN for Image Synthesis資料集DB
‘19-AAAI19/01/27Hierarchical Encoder with Auxiliary Supervision for Table-to-text Generation: Learning Better Representation for Tables表格轉文字
‘19-AAAI19/01/27A Radical-aware Attention-based Model for Chinese Text Classification漢字分類
‘19-CVPR19/02/25Handwriting Recognition in Low-resource Scripts using Adversarial Learning手寫辨識TF
‘19-CVPR19/03/27Tightness-aware Evaluation Protocol for Scene Text Detection評估CODE
‘19-ICCV19/05/31Scene Text Visual Question Answering資料集ICDAR_DB
‘19-CVPR19/06/16DynTypo: Example-based Dynamic Text Effects Transfer文字效果PRJ
VIDEO
‘19-CVPR19/06/16Typography with Decor: Intelligent Text Style Transfer文字效果*PYTORCH(M)
‘19-CVPR19/06/16An Alternative Deep Feature Approach to Line Level Keyword Spotting關鍵字檢索
‘19-ICCV19/07/23GA-DAN: Geometry-Aware Domain Adaptation Network for Scene Text Detection and Recognition領域適應
‘19-ICCV19/09/17Chinese Street View Text: Large-scale Chinese Text Reading with Partially Supervised Learning資料集ICDAR_DB
‘19-ICCV19/10/02Large-scale Tag-based Font Retrieval with Generative Feature Learning字型檢索
‘19-ICCV19/10/27TextPlace: Visual Place Recognition and Topological Localization Through Reading Scene Texts地點辨識DB
‘19-ICCV19/10/27DewarpNet: Single-Image Document Unwarping With Stacked 3D and 2D Regression Networks文件去扭曲*PYTORCH(M)
  • 本工作由 NAVER-LINE 旗下 Clova AI 的 OCR 團隊完成。NAVER-LINE 是亞洲頂尖的網路公司,並開發了雲端 AI 助理平台 Clova。
  • 本儲存庫預計配合主要 AI 會議的時程定期更新。

HagiCode

HagiCode 是面向專案開發的 AI 編碼工作台:豐富、便利的介面互動讓規劃、Agent 執行與專案工作更容易上手和跟進,帶來更投入的開發體驗。

透過豐富、便利的介面互動,讓規劃、Agent 執行與專案工作更容易上手和跟進。

造訪 HagiCode閱讀文件
  • 智慧OpenSpec 工作流程將意圖轉化為從提案、設計、任務到交付的可執行路徑。
  • 高效多 Agent 並行推進研究、實作與審查,並支援多儲存庫變更與你熟悉的 Agent CLI。
  • 有趣豐富、便利的介面互動讓規劃、Agent 執行與專案工作更容易上手和跟進,帶來更投入的開發體驗。

下載 Windows 版

從 Microsoft Store 取得所有下載

讓想法化為交付成果。

HagiCode 是面向專案開發的 AI 編碼工作台:以 OpenSpec 工作流程規劃變更,多 Agent 並行執行,再由 Hero Scenario 讓進度一目了然,將想法化為交付成果。

智慧
OpenSpec 工作流程將意圖轉化為從提案到交付的可執行路徑。
高效
多 Agent 並行推進研究、實作與審查。
有趣
豐富、便利的介面互動,讓規劃、Agent 執行與專案工作更容易跟進。

要離開 Awesome Site 嗎?

您即將前往外部網站。我們無法控制其內容或隱私慣例。

目的地

繼續前往網站