A Transformer-Based Framework for Scene Text Recognition

Text recognition in natural scene images is a challenging problem in computer vision. Different than the optical character recognition (OCR), text recognition in natural scene images is more complex due to variations in text size, colors, fonts, orientations, complex backgrounds, occlusion, illuminations and uneven lighting conditions. In this paper, we propose a segmentation-free method based on a deep convolutional recurrent neural network to solve the problem of cursive text recognition, particularly focusing on Urdu text in natural scenes. Compared to the non-cursive scripts, Urdu text recognition is more complex due to variations in the writing styles, several shapes of the same character, connected text, ligature overlapping, stretched, diagonal and condensed text. The proposed model gets a whole word image as an input without pre-segmenting into individual characters, and then transforms into the sequence of the relevant features. Our model is based on three components: a deep convolutional neural network (CNN) with shortcut connections to extract and encode the features, a recurrent neural network (RNN) to decode the convolutional features, and a connectionist temporal classification (CTC) to map the predicted sequences into the target labels. To increase the text recognition accuracy further, we explore deeper CNN architectures like VGG-16, VGG-19, ResNet-18 and ResNet-34 to extract more appropriate Urdu text features, and compare the recognition results. To conduct the experiments, a new large-scale benchmark dataset of cropped Urdu word images in natural scenes is developed. The experimental results show that the proposed deep CRNN network with shortcut connections outperform than other network architectures. The dataset is publicly available and can be downloaded from <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://data.mendeley.com/datasets/k5fz57zd9z/1</uri> .

Connectionist Temporal Classification (CTC) and attention mechanism are two main approaches used in recent scene text recognition works. Compared with attention-based methods, CTC decoder has a much shorter inference time, yet a lower accuracy. To design an efficient and effective model, we propose the guided training of CTC (GTC), where CTC model learns a better alignment and feature representations from a more powerful attentional guidance. With the benefit of guided training, CTC model achieves robust and accurate prediction for both regular and irregular scene text while maintaining a fast inference speed. Moreover, to further leverage the potential of CTC decoder, a graph convolutional network (GCN) is proposed to learn the local correlations of extracted features. Extensive experiments on standard benchmarks demonstrate that our end-to-end model achieves a new state-of-the-art for regular and irregular scene text recognition and needs 6 times shorter inference time than attention-based methods.

Scene text detection and recognition are vital challenging tasks in computer vision, which are to detect and recognize sequences of texts in natural scenes. Recently, researchers have investigated a lot of state-of-the-art methods to improve the accuracy and efficiency of text detection and recognition. However, there has been little research on text detection and recognition in natural scenes in Vietnam. In this paper, a deep learning-based empirical investigation of Vietnamese scene text detection and recognition is presented. Firstly, four detection models including differentiable binarization network (DBN), pyramid mask text detector (PMTD), pixel aggregation network (PAN), and Fourier contour embedding network (FCEN), are employed to detect text regions from the images. Then, four text recognition models including convolutional recurrent neural network (CRNN), self-attention text recognition network (SATRN), no-recurrence sequence-to-sequence text recognizer (NRTR), and RobustScanner (RS) are also investigated to recognize the texts. Moreover, data augmentation methods are also applied to enrich data for improving the accuracy and enhancing the performance of scene text detection and recognition. To validate the effectiveness of scene text detection and recognition models, the VinText dataset is employed for evaluation. Empirical results show that PMTD and SATRN achieve the highest scores among the others for text detection and recognition, respectively. For knowledge-sharing, our implementation is publicly available at https://github.com/ThorPham/VN_scene_text_detection_recognition.

Text recognition in the wild is a challenging task in the field of computer vision and machine learning. Existing optical character recognition engines cannot perform well in the natural scene. In this context, deep learning models have emerged as a powerful state-of-the-art technique in the classification and recognition process. This study proposes a new Convolutional Neural Network based system for scene text reading. We investigate how to combine the character recognition module followed by the word recognition module to achieve the overall system goal. The first module analyzes characters within multi-scale images by relaying on the power of the convolutional network and the fully connected network for character recognition. The second module relies on the Viterbi search to find the closest word to a given characters sequence. For the sake of more precision, a bigram based linguistic module is applied. The proposed system achieves the state-of-the-art performance on three standard scene text recognition benchmarks: chars74k, ICDAR 2003 and ICDAR 2013. In particular, this performance is proven on both of character and word recognition accuracy as well as speed aspects via a comparative study between different deep learning architectures.

Scene texts refer to arbitrary text found in images captured by cameras in real-world settings. The tasks of text detection and recognition are critical components of computer vision, with applications spanning scene understanding, information retrieval, robotics, and autonomous driving. Despite significant advancements in deep learning methods, achieving accurate text detection and recognition in complex images remains a formidable challenge for robust real-world applications. Several factors contribute to these challenges. First, the diversity of text shapes, fonts, colors, and styles complicates detection efforts. Second, the myriad combinations of characters, often with unstable attributes, make complete detection difficult, especially when background interruptions obscure character strokes and shapes. Finally, effective coordination of multiple sub-tasks in end-to-end learning is essential for success. This research aimed to tackle these challenges by enhancing text discriminative representation. This study focused on two interconnected problems: Scene Text Recognition (STR), which involves recognizing text from scene images, and Scene Text Detection (STD), which entails simultaneously detecting and recognizing multiple texts within those images. This research focuses on implementing and evaluating the Efficient and Accurate Scene Text Detector (EAST) algorithm for text detection and recognition in natural scene images. The study aims to compare the performance of three prominent Optical Character Recognition (OCR) techniques—TesseractOCR, PaddleOCR, and EasyOCR. The EAST model was applied to a series of sample test images, and the results were visually represented with bounding boxes highlighting the detected text regions. The inference times for each image were recorded, highlighting the algorithm's efficiency, with average times of 0.446, 0.439, and 0.440 seconds for the respective test images. These results indicate that the EAST algorithm is accurate and operates in real-time, making it suitable for applications requiring immediate text recognition.

Scene text detection is a crucial step in end-to-end scene text recognition, a greatly challenging problem in computer vision. This paper proposes a novel scene text detection method that involves superpixel-based stroke feature transform (SSFT) and deep learning based region classification (DLRC). The SSFT is developed for candidate character region (CCR) extraction, which consists in partitioning an input image into several regions via superpixel-based clustering, removing most regions based on predefined criteria satisfied by the characters, and refining the remaining regions to obtain CCRs by computing a stroke width map. The character regions are identified from the CCRs using DLRC, in which several hand-crafted low-level features, i.e., color, texture, and geometric features, and some deep convolution neural network (CNN) based high-level features are first extracted from the regions, and then these features are fused by using two fully connected networks (FCNs) for region classification. In the DLRC step, the deep feature extraction CNN and the feature fusion FCNs are jointly trained. Next, the extracted character regions are merged to form candidate text regions, from which the final scene texts are detected. The proposed method is evaluated on three publicly available datasets: ICDAR2011, ICDAR2013, and street view text. It achieves F -measures of 0.876, 0.885, and 0.631, respectively, which demonstrate the effectiveness of the proposed scene text detection method.

Scene text recognition (STR) involves reading text from images of natural scenes. The texts in such images come in a wide array of fonts, shapes, and orientations. Therefore, various works often rely on rectification network to rectify text images before passing them to the recognition network. However, rectifying an image that does not require it may create unwanted distortion. This may result in wrong predictions that would have otherwise been correct. In order to alleviate the adverse impact of rectification, a portmanteauing of features is presented. The method is introduced to a transformer-based model through a proposed block matrix initialisation which achieved competitive results. Although transformer has achieved notable success in various fields, areas for improvements with its application in STR were identified in this study. Firstly, vision transformer requires an input image to be resized into fixed height and width before being split into patches. However, it was discovered that certain patch resolutions seem to result in better accuracy for images with particular original aspect ratios. Secondly, the first decoded character generally has lower accuracy. In view of these issues, pure transformer with integrated experts (PTIE) is proposed. PTIE is able to process multiple patch resolutions and decode in both the original and reverse character orders thereby capitalising on the aforementioned areas and achieved state-of-the-art results. Handwritten text recognition (HTR) deals with handwritten text images that come from scanned or photograph documents. Works that employed transformer-based models often train them with additional synthetic data. However, these data are not publicly available. Furthermore, experimentation in this study seems to suggest that transformer trained on real HTR data generalises poorly to unseen data. Therefore, in the scope of real data, a simple transformer model which outperformed related works is presented in this thesis. This is achieved by adopting attention masking which addressed the issue of generalisation as well as introducing various pre-processing methods.

Building robust recognizers for Arabic has always been challenging. We demonstrate the effectiveness of an end-to-end trainable CNN-RNN hybrid architecture in recognizing Arabic text in videos and natural scenes. We outperform previous state-of-the-art on two publicly available video text datasets - ALIF and ACTIV. For the scene text recognition task, we introduce a new Arabic scene text dataset and establish baseline results. For scripts like Arabic, a major challenge in developing robust recognizers is the lack of large quantity of annotated data. We overcome this by synthesising millions of Arabic text images from a large vocabulary of Arabic words and phrases. Our implementation is built on top of the model introduced here [37] which is proven quite effective for English scene text recognition. The model follows a segmentation-free, sequence to sequence transcription approach. The network transcribes a sequence of convolutional features from the input image to a sequence of target labels. This does away with the need for segmenting input image into constituent characters/glyphs, which is often difficult for Arabic script. Further, the ability of RNNs to model contextual dependencies yields superior recognition results.

Scene text recognition is one of the most important techniques in pattern recognition and machine intelligence due to its numerous practical applications. Scene text recognition is also a sequence model task. Recurrent neural network (RNN) is commonly regarded as the default starting point for sequential models. Due to the non-parallel prediction and the gradient disappearance problem, the performance of the RNN is difficult to improve substantially. In this paper, a new TRDD network architecture which base on dilated convolution and residual block is proposed, using Convolutional Neural Networks (CNN) instead of RNN realizes the recognition task of sequence texts. Our model has the following three advantages in comparison to existing scene text recognition methods: First, the text recognition speed of the TRDD network is much fast than the state-of-the-art scene text recognition network based recurrent neural networks (RNN). Second, TRDD is easier to train, avoiding the problem of exploding and vanishing, which is major issue for RNN. Third, both using larger dilated factors and increasing the filter size are all viable ways to change receptive field size. We benchmark the TRDD on four standard datasets, it has higher recognition accuracy and faster recognition speed based on the smaller model. It is hopefully used in the real-time application.

Scene text recognition is a well-rooted research domain covering a diverse application area. Recognition of scene text is challenging due to the complex nature of scene images. Various structural characteristics of the script also influence the recognition process. Text and background segmentation is a mandatory step in the scene text recognition process. A text recognition system produces the most accurate results if the structural and contextual information is preserved by the segmentation technique. Therefore, an attempt is made here to develop a robust foreground/background segmentation(separation) technique that produces the highest recognition results. A ground-truth dataset containing Devanagari scene text images is prepared for the experimentation. An encoder-decoder convolutional neural network model is used for text/background segmentation. The model is trained with Devanagari scene text images for pixel-wise classification of text and background. The segmented text is then recognized using an existing OCR engine (Tesseract). The word and character level recognition rates are computed and compared with other existing segmentation techniques to establish the effectiveness of the proposed technique.

End-to-end Frameworks with Connectionist Temporal Classification (CTC) have achieved great success in text recognition. Despite high accuracies with deep learning, CTC-based text recognition methods also suffer from poor alignment (character boundary positioning) in many applications. To address this issue, we propose an end-to-end text recognition method based on robust prototype learning. In the new CTC framework, we formulate the blank as the rejection of character classes and use the one-vs-all prototype classifier as the output layer of the convolutional neural network. For network learning, based on forced alignment between frames and character labels, the most aligned frame is up-weighted in CTC training strategy to reduce estimation errors in decoding. Experiments of handwritten text recognition on four benchmark datasets of different languages show that the proposed method consistently improves the accuracy and alignment of CTC-based text recognition baseline.

A Metaverse text recognition model based on character-level contrastive learning

In this paper, we propose a new algorithm that conjointly address scene text detection and recognition by sharing convolutional feature map. Compared with most systems which consider text detection and recognition to be irrelevant tasks, we integrate text detection and recognition into an end-to-end trainable neural network based on convolutional and recurrent neural network It can precisely detect and recognize text during a simple forward propagation, avoiding redundant processes like image patch cropping, repeated calculation of feature map. We train this unified neural network just by images and corresponding ground truth bounding boxes and text labels. Our algorithm gains outstanding performance in terms of computation time and accuracy on standard benchmark datasets. The proposed model runs robustly on multi-ratios images without complicated post-processing steps.

In this paper, we present a DAV-based system for text (mainly English and Chinese) detection and recognition. With the combination of unmanned aerial vehicle and scene text recognition, the system can realize text detection and recognition in long-range air-plane images, providing an underlay for unmanned navigation and fast text information understanding. Robust text detection and accurate text recognition can be achieved by two contributions. First, a scalable engine is proposed to synthesize text images by overlaying English or Chinese text into existing images in a natural way. Second, an framework which is trainable and end-to-end by combining Convolutional Neural Network and Recurrent Neural Network is adapted to recognize the variable-length text with a high accuracy. Field experiments are performed with different videos shot in various backgrounds and outdoors to show that the proposed system can detect and recognise text information in UAV imagery robustly and effectively.

Although font recognition is a fundamental issue in the field of document analysis and recognition, it was usually ignored in the past. With the development of optical character recognition (OCR), font recognition becomes more and more important. This paper proposed a well-designed convolutional neural network (CNN) architecture for traditional Mongolian font recognition by means of a single word. To be specific, the whole word image is regarded as input of CNN. Hence, the word images should be normalized into the same size before being inputted into CNN. By comparison, an appropriate aspect ratio for the traditional Mongolian word images has been determined. Experimental results demonstrate that the proposed CNN architecture outperforms three classic CNN architectures, including LeNet-5, AlexNet and GoogLeNet. Therefore, the proposed CNN is much more suitable for the task of the traditional Mongolian font recognition in the way of a single word.