scispace - formally typeset
Search or ask a question
Journal ArticleDOI

Offline Handwritten Devanagari Word Recognition Using CNN-RNN-CTC

13 Dec 2022-SN computer science-Vol. 4, Iss: 1, pp 1-12
About: This article is published in SN computer science.The article was published on 2022-12-13. It has received 2 citations till now. The article focuses on the topics: Computer science & Devanagari.
Citations
More filters
Proceedings ArticleDOI
17 Mar 2023
TL;DR: In this article , the authors proposed a novel framework that takes curvature into account when learning and employs CNN LSTM and Yolov5-OCR to extract information from printed and handwritten documents.
Abstract: The field of automated cheque collection has received a lot of attention recently due to the complexity and time-consuming nature of the manual process, which requires significant investment in technology, infrastructure, and staff training. However, some key areas such as Customer Experience and Interoperability with Image Processing and machine learning systems have not been explored. The challenge lies in the fact that handwritten characters are unique, and there is a lack of open-sourced foundation models. Furthermore, cheques and bank receipts contain both handwritten and non-handwritten text, making it difficult to extract information. The research proposes a novel framework that takes curvature into account when learning and employs CNN LSTM and Yolov5-OCR to extract information from printed and handwritten documents. The CTC-loss-function has limited application to processing bank cheques we had identified this gap and further improved overfitting which results due to CTC loss by combining CTC, curvature loss, and embedding loss, This research has further contributed to the development of an automated cheque collection framework that can extract information from both printed and handwritten texts, improving the overall efficiency of the cheque collection process and interoperability within various systems which is laborious in nature.
Proceedings ArticleDOI
17 Mar 2023
TL;DR: In this article , the authors proposed a novel framework that takes curvature into account when learning and employs CNN LSTM and Yolov5-OCR to extract information from printed and handwritten documents.
Abstract: The field of automated cheque collection has received a lot of attention recently due to the complexity and time-consuming nature of the manual process, which requires significant investment in technology, infrastructure, and staff training. However, some key areas such as Customer Experience and Interoperability with Image Processing and machine learning systems have not been explored. The challenge lies in the fact that handwritten characters are unique, and there is a lack of open-sourced foundation models. Furthermore, cheques and bank receipts contain both handwritten and non-handwritten text, making it difficult to extract information. The research proposes a novel framework that takes curvature into account when learning and employs CNN LSTM and Yolov5-OCR to extract information from printed and handwritten documents. The CTC-loss-function has limited application to processing bank cheques we had identified this gap and further improved overfitting which results due to CTC loss by combining CTC, curvature loss, and embedding loss, This research has further contributed to the development of an automated cheque collection framework that can extract information from both printed and handwritten texts, improving the overall efficiency of the cheque collection process and interoperability within various systems which is laborious in nature.
References
More filters
Proceedings Article
07 Dec 2015
TL;DR: This work introduces a new learnable module, the Spatial Transformer, which explicitly allows the spatial manipulation of data within the network, and can be inserted into existing convolutional architectures, giving neural networks the ability to actively spatially transform feature maps.
Abstract: Convolutional Neural Networks define an exceptionally powerful class of models, but are still limited by the lack of ability to be spatially invariant to the input data in a computationally and parameter efficient manner. In this work we introduce a new learnable module, the Spatial Transformer, which explicitly allows the spatial manipulation of data within the network. This differentiable module can be inserted into existing convolutional architectures, giving neural networks the ability to actively spatially transform feature maps, conditional on the feature map itself, without any extra training supervision or modification to the optimisation process. We show that the use of spatial transformers results in models which learn invariance to translation, scale, rotation and more generic warping, resulting in state-of-the-art performance on several benchmarks, and for a number of classes of transformations.

6,150 citations

Journal ArticleDOI
TL;DR: Zhang et al. as mentioned in this paper proposed a novel neural network architecture, which integrates feature extraction, sequence modeling and transcription into a unified framework, and achieved remarkable performances in both lexicon free and lexicon-based scene text recognition tasks.
Abstract: Image-based sequence recognition has been a long-standing research topic in computer vision. In this paper, we investigate the problem of scene text recognition, which is among the most important and challenging tasks in image-based sequence recognition. A novel neural network architecture, which integrates feature extraction, sequence modeling and transcription into a unified framework, is proposed. Compared with previous systems for scene text recognition, the proposed architecture possesses four distinctive properties: (1) It is end-to-end trainable, in contrast to most of the existing algorithms whose components are separately trained and tuned. (2) It naturally handles sequences in arbitrary lengths, involving no character segmentation or horizontal scale normalization. (3) It is not confined to any predefined lexicon and achieves remarkable performances in both lexicon-free and lexicon-based scene text recognition tasks. (4) It generates an effective yet much smaller model, which is more practical for real-world application scenarios. The experiments on standard benchmarks, including the IIIT-5K, Street View Text and ICDAR datasets, demonstrate the superiority of the proposed algorithm over the prior arts. Moreover, the proposed algorithm performs well in the task of image-based music score recognition, which evidently verifies the generality of it.

2,184 citations

Journal ArticleDOI
TL;DR: In this article, a new adaptation layer is proposed to reduce the mismatch between training and test data on a particular source layer, and the adaptation process can be efficiently and effectively implemented in an unsupervised manner.

232 citations

Journal ArticleDOI
01 Nov 2011
TL;DR: In this paper, the state of the art from 1970s of machine printed and handwritten Devanagari optical character recognition (OCR) is discussed in various sections of the paper.
Abstract: In India, more than 300 million people use Devanagari script for documentation. There has been a significant improvement in the research related to the recognition of printed as well as handwritten Devanagari text in the past few years. State of the art from 1970s of machine printed and handwritten Devanagari optical character recognition (OCR) is discussed in this paper. All feature-extraction techniques as well as training, classification and matching techniques useful for the recognition are discussed in various sections of the paper. An attempt is made to address the most important results reported so far and it is also tried to highlight the beneficial directions of the research till date. Moreover, the paper also contains a comprehensive bibliography of many selected papers appeared in reputed journals and conference proceedings as an aid for the researchers working in the field of Devanagari OCR.

159 citations

Journal ArticleDOI
TL;DR: Various feature extraction and classification techniques associated with the offline handwriting recognition of the regional scripts are discussed in this survey, which will serve as a compendium not only for researchers in India, but also for policymakers and practitioners in India.
Abstract: Offline handwriting recognition in Indian regional scripts is an interesting area of research as almost 460 million people in India use regional scripts. The nine major Indian regional scripts are Bangla (for Bengali and Assamese languages), Gujarati, Kannada, Malayalam, Oriya, Gurumukhi (for Punjabi language), Tamil, Telugu, and Nastaliq (for Urdu language). A state-of-the-art survey about the techniques available in the area of offline handwriting recognition (OHR) in Indian regional scripts will be of a great aid to the researchers in the subcontinent and hence a sincere attempt is made in this article to discuss the advancements reported in this regard during the last few decades. The survey is organized into different sections. A brief introduction is given initially about automatic recognition of handwriting and official regional scripts in India. The nine regional scripts are then categorized into four subgroups based on their similarity and evolution information. The first group contains Bangla, Oriya, Gujarati and Gurumukhi scripts. The second group contains Kannada and Telugu scripts and the third group contains Tamil and Malayalam scripts. The fourth group contains only Nastaliq script (Perso-Arabic script for Urdu), which is not an Indo-Aryan script. Various feature extraction and classification techniques associated with the offline handwriting recognition of the regional scripts are discussed in this survey. As it is important to identify the script before the recognition step, a section is dedicated to handwritten script identification techniques. A benchmarking database is very important for any pattern recognition related research. The details of the datasets available in different Indian regional scripts are also mentioned in the article. A separate section is dedicated to the observations made, future scope, and existing difficulties related to handwriting recognition in Indian regional scripts. We hope that this survey will serve as a compendium not only for researchers in India, but also for policymakers and practitioners in India. It will also help to accomplish a target of bringing the researchers working on different Indian scripts together. Looking at the recent developments in OHR of Indian regional scripts, this article will provide a better platform for future research activities.

133 citations