TextOCR: Towards large-scale end-to-end reasoning for arbitrary-shaped scene text

Singh, Amanpreet; Pang, Guangchang; Toh, Mandy; Huang, Jing; Galuba, Wojciech; Hassner, Tal

doi:10.1109/cvpr46437.2021.00869

Cited by 100 publications

(53 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although there are many data sets for training and evaluating OCR systems on printed and scene text (see Singh et al for a recent survey [42]), we found that existing collections of handwritten word images to be limited in their variability (See Table 1). To facilitate training and testing of our TSB, we, therefore, collected and annotated ∼135K handwritten English words from 5K images originally hosted publicly on Imgur.com.…”

Section: Imgur5k Handwriting Setmentioning

confidence: 89%

TextStyleBrush: Transfer of Text Aesthetics from a Single Example

Krishnan¹,

Kovvuri²,

Pang³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

We present a novel approach for disentangling the content of a text image from all aspects of its appearance. The appearance representation we derive can then be applied to new content, for one-shot transfer of the source style to new content. We learn this disentanglement in a self-supervised manner. Our method processes entire word boxes, without requiring segmentation of text from background, per-character processing, or making assumptions on string lengths. We show results in different text domains which were previously handled by specialized methods, e.g., scene text, handwritten text. To these ends, we make a number of technical contributions: (1) We disentangle the style and content of a textual image into a non-parametric, fixed-dimensional vector. (2) We propose a novel approach inspired by StyleGAN but conditioned over the example style at different resolution and content. (3) We present novel self-supervised training criteria which preserve both source style and target content using a pre-trained font classifier and text recognizer. Finally, (4) we also introduce Imgur5K, a new challenging dataset for handwritten word images. We offer numerous qualitative photo-realistic results of our method. We further show that our method surpasses previous work in quantitative tests on scene text and handwriting datasets, as well as in a user study.

show abstract

Section: Imgur5k Handwriting Setmentioning

confidence: 89%

TextStyleBrush: Transfer of Text Aesthetics from a Single Example

Krishnan¹,

Kovvuri²,

Pang³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The experimental results show that the proposed method has improvement on accuracy compared to previous text detectors, and achieves state-of-the-art performance on three benchmark datasets. In the future, we are interested in detecting cases "text inside text" via RCNN module and developing an end-to-end [16] scene text spotting system.…”

Section: Discussionmentioning

confidence: 99%

Real-Time Scene Text Detection Based on Global Level and Word Level Features

Zhao¹,

Yu²,

Xing³

et al. 2022

SSRN Journal

View full text Add to dashboard Cite

“…Having in mind the idea that the text recognition head might be undertrained due to complex training pipeline, we freeze all layers except layers that are related to the text recognition head and fine-tune the model during 130K iterations. This fine-tuning does not improve results on ICDAR 2015, but improves quality results upto 74.5% End-to-end recognition metric on Total-Text dataset, making quality metrics values on par with Mask TextSpotter v3 fine-tuned on TextOCR dataset (Singh et al (2021)), see Table 5 for details.…”

Section: Methodsmentioning

confidence: 99%

“…TextOCR (Singh et al (2021)) is a recently published arbitrary-shaped scene text detection and recognition dataset consisting of 28,134 images with 900K annotated words.…”

Section: Scene Text Datasetsmentioning

confidence: 99%

Open Images V5 Text Annotation and Yet Another Mask Text Spotter

Krylov¹,

Nosov²,

Sovrasov³

2021

Preprint

View full text Add to dashboard Cite

A large scale human-labeled dataset plays an important role in creating high quality deep learning models. In this paper we present text annotation for Open Images V5 dataset. To our knowledge it is the largest among publicly available manually created text annotations. Having this annotation we trained a simple Mask-RCNN-based network, referred as Yet Another Mask Text Spotter (YAMTS), which achieves competitive performance or even outperforms current state-of-the-art approaches in some cases on ICDAR 2013, ICDAR 2015 and Total-Text datasets. Code for text spotting model available online at: https://github.com/openvinotoolkit/training_extensions. The model can be exported to OpenVINO™-format and run on Intel® CPUs.

show abstract

TextOCR: Towards large-scale end-to-end reasoning for arbitrary-shaped scene text

Cited by 100 publications

References 35 publications

TextStyleBrush: Transfer of Text Aesthetics from a Single Example

TextStyleBrush: Transfer of Text Aesthetics from a Single Example

Real-Time Scene Text Detection Based on Global Level and Word Level Features

Open Images V5 Text Annotation and Yet Another Mask Text Spotter

Contact Info

Product

Resources

About