From object detection to text detection and recognition: A brief evolution history of optical character recognition

Author(s):  
Haifeng Wang ◽  
Changzai Pan ◽  
Xiao Guo ◽  
Chunlin Ji ◽  
Ke Deng
2020 ◽  
Vol 2020 (1) ◽  
pp. 78-81
Author(s):  
Simone Zini ◽  
Simone Bianco ◽  
Raimondo Schettini

Rain removal from pictures taken under bad weather conditions is a challenging task that aims to improve the overall quality and visibility of a scene. The enhanced images usually constitute the input for subsequent Computer Vision tasks such as detection and classification. In this paper, we present a Convolutional Neural Network, based on the Pix2Pix model, for rain streaks removal from images, with specific interest in evaluating the results of the processing operation with respect to the Optical Character Recognition (OCR) task. In particular, we present a way to generate a rainy version of the Street View Text Dataset (R-SVTD) for "text detection and recognition" evaluation in bad weather conditions. Experimental results on this dataset show that our model is able to outperform the state of the art in terms of two commonly used image quality metrics, and that it is capable to improve the performances of an OCR model to detect and recognise text in the wild.


Author(s):  
Andrew Brock ◽  
Theodore Lim ◽  
J. M. Ritchie ◽  
Nick Weston

End-to-end machine analysis of engineering document drawings requires a reliable and precise vision frontend capable of localizing and classifying various characters in context. We develop an object detection framework, based on convolutional networks, designed specifically for optical character recognition in engineering drawings. Our approach enables classification and localization on a 10-fold cross-validation of an internal dataset for which other techniques prove unsuitable.


Author(s):  
Zhang Yun-An ◽  
Pan Ziheng ◽  
Dui Hongyan ◽  
Bai Guanghan

Background: YOLOv3-Tesseract is widely used for the intelligent form recognition because it exhibits several attractive properties. It is important to improve the accuracy and efficiency of the optical character recognition. Methods: The YOLOv3 exhibits the classification advantages for the object detection. Tesseract can effectively recognize regular characters in the field of the optical character recognition. In this study, a YOLOv3 and Tesseract-based model of improved intelligent form recognition is proposed. Results: First, YOLOv3 is trained to detect the position of the text in the table and to subsequently segment text blocks. Second, Tesseract is used to individually detect separated text blocks and combine YOLOv3 and Tesseract to achieve the goal of table character recognition. Conclusion: Based on the Tianchi big data, experimental simulation is used to demonstrate the proposed method. The YOLOv3-Tesseract model is trained and tested to effectively accomplish the recognition task.


Author(s):  
Yohei Igarashi

Although Coleridge is mostly known for being a copious talker who was impossible to transcribe, this chapter recovers Coleridge’s role as transcriber, theorist of transcription practices, and inventor of his own idiosyncratic shorthand. Considering Coleridge’s time as a parliamentary reporter, his self-reflexive notebook entries, and the history of stenography, this chapter posits that Coleridge pursued an efficient writing system to record not speech but the flow of his own silent thoughts. Also discussing today’s optical character recognition software and the shorthand effect (when letters or words uncannily become illegible shapes, and non-linguistic shapes come to look like linguistic signs), this chapter culminates in a reading of the “signs” in “The Rime of the Ancient Mariner.”


2021 ◽  
Vol 14 (4) ◽  
pp. 11
Author(s):  
Kayode David Adedayo ◽  
Ayomide Oluwaseyi Agunloye

License plate detection and recognition are critical components of the development of a connected Intelligent transportation system, but are underused in developing countries because to the associated costs. Existing license plate detection and recognition systems with high accuracy require the usage of Graphical Processing Units (GPU), which may be difficult to come by in developing nations. Single stage detectors and commercial optical character recognition engines, on the other hand, are less computationally expensive and can achieve acceptable detection and recognition accuracy without the use of a GPU. In this work, a pretrained SSD model and a tesseract tessdata-fast traineddata were fine-tuned on a dataset of more than 2,000 images of vehicles with license plate. These models were combined with a unique image preprocessing algorithm for character segmentation and tested using a general-purpose personal computer on a new collection of 200 automobiles with license plate photos. On this testing set, the plate detection system achieved a detection accuracy of 99.5 % at an IOU threshold of 0.45 while the OCR engine successfully recognized all characters on 150 license plates, one character incorrectly on 24 license plates, and two or more incorrect characters on 26 license plates. The detection procedure took an average of 80 milliseconds, while the character segmentation and identification stages took an average of 95 milliseconds, resulting in an average processing time of 175 milliseconds per image, or 6 photos per second. The obtained results are suitable for real-time traffic applications.


1997 ◽  
Vol 9 (1-3) ◽  
pp. 58-77
Author(s):  
Vitaly Kliatskine ◽  
Eugene Shchepin ◽  
Gunnar Thorvaldsen ◽  
Konstantin Zingerman ◽  
Valery Lazarev

In principle, printed source material should be made machine-readable with systems for Optical Character Recognition, rather than being typed once more. Offthe-shelf commercial OCR programs tend, however, to be inadequate for lists with a complex layout. The tax assessment lists that assess most nineteenth century farms in Norway, constitute one example among a series of valuable sources which can only be interpreted successfully with specially designed OCR software. This paper considers the problems involved in the recognition of material with a complex table structure, outlining a new algorithmic model based on ‘linked hierarchies’. Within the scope of this model, a variety of tables and layouts can be described and recognized. The ‘linked hierarchies’ model has been implemented in the ‘CRIPT’ OCR software system, which successfully reads tables with a complex structure from several different historical sources.


Sign in / Sign up

Export Citation Format

Share Document