Here is Your Sample Download Sample 📩
Image Processing and Analytics
Building density and manhole placement are two variables that the author has identified through statistical studies as they relate to manhole management. In order to find trends in the monthly and yearly amounts of rainfall, the study also applies image processing techniques to map data.
The study did note that the Python programming language with associated image processing libraries, like OpenCV, have been utilized to apply image processing techniques to map data. Popular open-source based image processing software package Open Source Computer Vision package) offers a number of image processing methods, including image filtering, feature identification, object recognition, and segmentation (Thakur, 2021).
A combination of statistical analysis, image processing methods, literature reviews, and extrapolation have been employed in the study to pinpoint the variables that affect manhole management and to suggest improvements for public utility networks in Indian towns. Although the accuracy of these methods and the quality of the data collected will determine how well they perform, they can offer insightful information on manhole management.
The research explains how CNES creates value-added products with the help of satellite data using a variety of techniques. The techniques employed comprise the creation of a reference framework as well as the processing and analysis of data using cloud technologies and big data. In order to manage the growing volume of data and processing requests and other information CNES has switched into another framework which is focused on cloud technologies and big data (Baillarin, et al., 2019).
This new framework most likely uses a variety of data processing and analysis algorithms to produce value-added products for monitoring disasters and the environment. It is conceivable that CNES's emphasis on employing big data and cloud technologies to handle and analyze satellite data will result in value-added products that are more precise and timelier.
An automated method can simplify the process of identifying cancer cells in their early stages. The study paper proposes an automated approach for leukemia identification utilizing image analytics and classification algorithms applied to patient cell image data. Two algorithms are mentioned in the research paper: both neural networks and K-means clustering. During the segmentation stage, the K-means clustering technique is employed to group the picture samples according to their shared features (Belhekar, Gagare, Bedse, & Bhelkar, 2019).
The classification step, which divides the segmented images into groups like normal and pathological cells, uses a neural networks method. The AUC or area under the curve for proposed automated system, which used image analytics and classification algorithms, was 0.865. Its calculation accuracy, precision, and F1 score for neural networks were also 0.838, 0.838, and 0.835, respectively. This shows that the algorithm was able to identify leukemia cells with a fair amount of accuracy in the collection of cell images that was made available.
The techniques employed in this research include the development of a novel convolution neural network (CNN) is known as workpiece-net or wp-net for item classification, as well as image processing technologies as canny edge detection operator, morphological processing, and de-noising processing (Chen, Lei, Liu, Zhang, & Yao, 2020). For object categorization, the study used a convolution neural network (CNN) called the workpiece-net (wp-net). The algorithm that underlies workpiece detection is this one. The research achieved an average Intersection over Union (IoU) of 0.9235, an average precision of 0.9883, and an accuracy of 0.9986 for image cutting. Without a graphics processing unit (GPU) or multithreading, classification is performed at a speed of 0.1243s/fps with an average recall of 0.9877.
2. Image Augmentation and Feature Extraction
The main focus of this study was to overcome the challenge of insufficient training data in deep learning for object classification in the manufacturing field. The reason for this challenge is that most objects within manufacturing field are often uncommon, and not associates with the existing datasets (Kim, Kim, Kim, & Kim, 2022). Overall, the study aimed to develop a methodology that could overcome the challenge of insufficient training data for deep learning in the manufacturing field by proposing a data augmentation technique based on color perturbation and feature pre-extraction.
Data augmentation and a deep learning model called Mask RCNN-FPN were used in this study to find and extract information from borderless tables in picture documents. The deep learning algorithm employed in this study is the Python-created Mask RCNN-FPN model. The data collected from the data augmentation block is processed using Python (Zulkarnain, Nurmalasari, & Azizah, 2022). Metrics like precision, recall, the F-measure, and intersection value were used in the research to assess how well the deep learning model performed in detecting and recognizing the structure of borderless tables.
The performance of the model was enhanced by the use of data augmentation with the help of the CutMask technique, although the results for table structure recognition did not compare favorably with those from earlier studies.
Image categorization, deep convolution neural networks (CNN), and data augmentation using geometric transformation and distortion injection are the techniques being deployed.
Multi-Stage Image Data Augmentation for Deep CNN includes the following steps
Upload the dataset of photographs of endangered and rare orchids in Indonesia. Separate the dataset into sets for training and validation. Use geometric transformation and distortion injection as part of a multi-stage picture data augmentation on the training dataset. Use ResNet as the CNN model and transfer learning (Dewantara, Hidayat, Susanto, & Arymurthy, 2020) .Use the increased training dataset to train the model. Assess the model's performance on the validation set. Evaluate the model using a different test dataset. Compare the outcomes of the suggested method with those of other methods. Using a Deep Convolution Neural Network to Classify Images involve the following steps.
Upload the dataset of photographs of endangered and rare orchids in Indonesia. Separate the dataset into sets for training and validation. Classify images using deep convolution neural networks (CNN). Examine the CNN model's performance on the validation set. Use a different test dataset to evaluate the CNN model. Compare the CNN model's outcomes to those of other classification techniques. ResNet with Transfer Learning involves the following steps. Load the ResNet trained model first. Add a new classification layer after the final completely connected layer.
Freeze the pre-trained layers' weights. Use the training dataset to train the new classification layer. Evaluate the model's performance on the validation set. Evaluate the model using a different test dataset. Examine the outcomes of the ResNet model in comparison to other models. The results of the research show that the proposed multi-stage image data augmentation approach has a convincing performance compared to existing methods.
The task of image classification process undergoes by analyzing maximum labeled training images to achieve high accuracy and generalization. However, manually labeling images is a time-consuming and expensive process. In addition, it can be challenging to collect sufficient training data in scenarios where only one training sample is available for each class. To overcome these challenges, this paper proposes a novel data augmentation technique that is specifically designed for color texture classification.
The proposed method uses color features information extracted from various color spaces to augment data when one training sample is given to individual class (Duong & Hoang, 2019). It demonstrates the potential to significantly improve classification performance using only a single training sample per class. This could have important implications for real-world applications, where collecting labeled training data can be difficult or costly.
The extraction of physical-related information from each simulated SAR images in data augmentation in SAR ATR applications, the research paper's solution uses an adversarial encoding network. It has been demonstrated that the strategy works well for addressing data scarcity issues and enhancing classification accuracy. The adversarial encoding network for synthetic aperture radar automated target recognition (SAR ATR) application is the algorithm that is covered in the passage (Du, Hong, Wang, Xing, & Qiu, 2021).
The proposed model can reach 98.55% accuracy, according to the paper, especially when there is not enough real data to make a classification. This shows that the adversarial encoding network suggested in the research is successful in extracting physically relevant characteristics from simulated SAR images, which can be used for data augmentation and enhance the functionality of SAR ATR applications.
3. Image Recognition
PCA can efficiently help to understand the features by scanning its representation and face recognition. feature vectors require more coefficients compared to PCA or Principal Component Analysis.The 2DPCA and Image PCA algorithms are the two that were employed in this paper. Principal component analysis in two dimensions (2DPCA): It is a method for facial recognition feature extraction. Compared to principal component analysis (PCA) with the help of 2D matrices and using multiple coefficients for feature vectors. Image PCA: The approach put forward in the paper entails using the projective feature image that was produced after the source photos were subjected to PCA. It is used to effectively portray and recognize faces (Wen & Shi, 2007).
The experimental findings demonstrate that, while requiring fewer coefficients for feature vectors than 2DPCA, image PCA achieves a similar or higher identification rate. Two face image databases, the Yale and ORL face databases are used in a number of experiments to test and evaluate image PCA. The results show that image PCA achieves a recognition rate is similar or sometimes much higher than that of 2DPCA while using fewer coefficients for feature vectors. PCA is therefore more accurate than 2DPCA.
The primary purpose of this research paper is to explores multiple applications under deep learning, specifically classification neural networks considering computer vision as a field for the classification of Tibetan ancient books and characters. The study employs four popular deep learning models, namely RestNet, VGGNet, LeNet, and Wide-ResNet, which help in image classification. The aim of the study is to compare and analyze the results obtained from these models (Zhao, 2022). The Wide-ResNet model has a direct effect on the classification of Tibetan images. It is nearly 94% accurate.
In order to identify and categories diseases using chest X-rays, the research outlines the creation of an automated chest X-ray classification method using the Manta Ray Optimization using MRFO-DLA or Deep Learning Approach. The method uses an MRFO with an autoencoder (AE) model for CXR classification, a neural architecture search network (NASNet) in order to apply feature extraction, and the bilateral filtering (BF) methodology for image preprocessing. Bilateral filtering (BF), neural architecture search network (NASNet), and MRFO with autoencoder (AE) models are all used in the MRFO-DLA method (Kumar & Ponnusamy, 2022).
Image preprocessing is done with BF, feature extraction is done with NASNet, and CXR classification is done with MRFO and an AE model. The performance of the classification models is measured throughout the study using the accuracy rate (Kumar & Ponnusamy, 2022). With an accuracy rate of roughly 94%, the Wide-ResNet model can provide the best classification impact by analyzing Tibetan single-character images.
By leveraging deep learning techniques and CNNs specifically, this study aims to develop an approach that is both effective and efficient for classifying fireworks images. According to th research paper, the findings will create an important impact on the applications that require accurate image recognition, such as in the fields of security, surveillance, and photography. By demonstrating the effectiveness of deep learning techniques and CNNs for image classification, this study provides a foundation for further research and development in this important area.
In order to appropriately identify and classify apparel photos, the researchers used machine learning (ML) and deep learning (DL) techniques. Support vector machines (SVM), Decision Tree (DT), K-Nearest Neighbours (KNN), and Random Forest (RF) were the machine learning algorithms employed. Convolution Neural Networks (CNN), GoogleNet, AlexNet, LeNet were the DL methods used (Samia, Soraya & Malika, 2022). Additionally, transfer learning was carried out utilizing the trained models VGG16, MobileNet, and RestNet50.
The ANN model produced meaningful results for machine learning models, with an accuracy of 88.71%. The GoogleNet architecture produced impactful results for DL models, with an accuracy of 93.75%. Accuracy and matrix confusion were the primary metrics utilized to assess how well ML and DL algorithms performed. According to the study, getting the best outcomes depends on the network's depth and the number of epochs.
Magnetic Resonance Imaging (MRI) is a widely used scanning technique to visualize brain tumors. By processing images collected from MRI scans using deep learning techniques, neurologists can more accurately classify brain tumors. An exploratory analysis in this paper focuses on of brain MRI images based on extracted features and a comparative analysis of different Convolution Neural Network (CNN) based transfer learning models for the classification of MRI images for brain tumors (Arora & Sharma, 2021).
The ultimate goal of this research is maintain the accuracy level in the tumor detection process of the brain and also classify them using a combination of pattern analysis, medical image processing, segmentation, and computer vision techniques for enhancement, and classification of brain diagnosis. Using deep learning methods for the classification of brain tumors in MRI images offers several advantages over traditional approaches. The extracted features allow for a more nuanced and accurate analysis of the images, while the use of transfer learning models enables a more efficient and effective analysis of the data.
4. Sign Language detection
4.1 Deep Learning
The most popular language to express feelings infront of others for deafend communities is sign language. Because of the complexities of this language, it is difficult for the general public to understand and communicate with the deaf and hard of hearing community. The suggested method detects real-time sign language using a customized CNN model trained on a dataset including 11 sign words. The dataset was preprocessed before training the CNN model to improve model accuracy. The findings demonstrate that the suggested customized CNN model on the test dataset obtained high accuracy of 98.6%, as well as high precision, f1-score, recall, and of 99% (Shahriar, et al., 2018). By enabling real-time sign language detection, the research's findings have the potential to have a significant influence on communication between hearing and hearing-impaired groups.
A popular method of visual communication among Bangladesh's hearing-impaired population is Bangladeshi Sign Language (BdSL). Real-time BdSL sign recognition from photos is a difficult task. This study suggests a novel method for instantly identifying BdSL indications in photos (Hoque, Jubair, Islam, Akash, & Paulson, 2018). The technique recognizes the classes of signs present in the image region using a Convolution Neural Network (CNN) based object detection technique. Previous research on detecting BdSL indicators has been hampered by reliance on external sensors, and other vision-based algorithms have not achieved efficient real-time performance. In contrast, the suggested technique overcomes such constraints and provides successful real-time identification and recognition of Bangladeshi signs.
Personal and professional development requires effective communication. Due to communication hurdles caused by their inability to use spoken language, people with speak and hearing disabilities are unable to communicate effectively with others who do not have such disabilities (Jindal, Yadav, Nirvan, & Kumar, 2022). In India, disabled persons account for 2.21% of the total population, with 19% having hearing impairment and 7% having speech impairment. Sign language movements, which are frequently employed by people with hearing impairments, are frequently difficult for others to interpret.
To address this issue, this study used AlexNet in MATLAB and Convolution Neural Network (CNN) in Python and to develop two models for converting sign movements to text. The suggested CNN and AlexNet models use image processing techniques to recognize and convert gestures performed by people with hearing difficulties. The proposed models have demonstrated encouraging results in recognizing sign language movements, and they have the potential to make communication more accessible and effective for those with hearing or speech problems and those without such disabilities.
Sign language recognition is an important task that can assist people with hearing and speech disabilities in communicating effectively. A revolutionary deep multi-layered convolution neural network is proposed in this research report to automatically detect and classify sign language from hand motion photographs. The suggested method makes use of a deep CNN structure made up of 32 convolution filters with a 3 x 3 kernel, a LeakyReLU activation function, and a 2 x 2 max pooling operation. In addition, in the output layer, a SoftMax activation function was used to categorize the sign language gesture (Bhadra & Kar, 2021). A database of static (54,000 photos and 36 classes) and dynamic (49,613 images and 23 classes) hand gesture images was utilized to test the performance of the suggested approach.
Experimental results reveal that the proposed approach performs exceptionally well in the sign language detection task. The efficacy of the proposed approach can be attributed to the use of a deep multi-layered CNN structure that can learn and extract robust features from the hand gesture images. Overall, the proposed approach offers a promising solution for the automatic detection and classification of sign language from hand gesture images, which can help individuals with hearing and speech impairments communicate more effectively.
Expressing oneself is essential to convey ideas and emotions. However, nonverbal autism can make it difficult for some individuals to communicate verbally. Therefore, sign language has become an alternative mode of communication for speech-impaired individuals. However, sign language can be difficult for non-signers to understand. In this paper, we present a model that aims to bridge the communication gap between speech-impaired individuals and non-signers. The proposed model converts sign language gestures into text and further into audio format.
The model uses a video input from the camera, and AlexNet, which the Convolution Neural Network architecture, to classify real-time hand gestures (Unkule, Shinde, Saurkar, Agarkar, & Verma, 2022). During training, the model was trained on more than 400 images of 10 gestures consisting of alphabets, numbers, and phrases. The images were preprocessed with canny edge filtering to improve accuracy. The recognized gesture is then converted into audio. The proposed model achieved an accuracy of 95.31 % with a 9:1 training-testing ratio. The proposed model enables individuals with speech impairments to communicate with non-signers in a more effective manner.
Effective communication is crucial for interpersonal relationships, but it can be difficult for those who have speech or hearing impairments. The main form of communication for the deaf and dumb is sign language. Communication between hearing-impaired people and non-signers is made possible by understanding sign language (Htet, Aung, & Aye, 2022). An object detection model based on YOLO, a classification model based on CNN, and a recognition model based on graphs have been built as part of a vision-based sign language recognition system in this context.
The technology is built to accurately and instantly recognize 29 different Myanmar sign language hand signs with a 98% accuracy rate. The experiment's findings demonstrate how effectively the suggested method can quickly and precisely identify Myanmar sign language. With the aid of this method, hearing-impaired individuals and non-signers can effectively communicate with each other by utilizing sign language.
4.2 Transfer Learning
This study focuses on creating the best method for instantly understanding Bangla Sign language (BdSL), which is widely used by Bangladesh's deaf and dumb population. The suggested method calls for the creation of a brand-new dataset named BdSLInfinite, which comprises of 2,000 pictures of 37 various signs. After that, a convolution neural network (CNN) based model using the Xception architecture is trained using the dataset. With an outstanding 98.93% accuracy over the test set and an average response time of 48.53 ms, the model performs admirably.
These findings show that, in terms of accuracy and speed, the proposed strategy greatly outperforms all currently used BdSL identification techniques (Urmee, Al Mashud, Akter, & Islam, 2019). The outcomes of the suggested method show that it has the potential to be applied in situations where it is important to communicate with the deaf and dumb community. Therefore, by giving the deaf and dumb community in Bangladesh a more effective and dependable communication infrastructure, the suggested solution can significantly improve their communication experience.
People with speech and hearing problems are increasingly using sign language as a means of communication. Given the potential uses in a variety of industries, including gaming, robotics, and security access control, sign language recognition has emerged as an active study area. In this research, a Python-based GUI that can employ a trained model to recognize hand gestures as a security access control technique on a single frame from a camera is provided. After 14 epochs, the proposed model achieves a layer loss of 2.803 and a mAP of 98.69%, showing that it is highly accurate at identifying hand movements. (Susa, Macalisang, Sevilla, Evangelista, Quismundo, & Reyes, 2022) The study shows that this model performs better in terms of validation accuracy than earlier comparable research.
Sign language recognition through computer vision has numerous practical applications, including the authentication of security access control systems. The proposed model uses a deep learning-based approach to detect hand gestures in real-time, making it suitable for security purposes (Susa, et al. 2022). The study is conducted by creating a Python-based graphical user interface (GUI) that processes a single frame from a camera to recognize the gestures. The model's accuracy is measured in terms of mean average precision (mAP) and layer loss. The experimental results show that the proposed method can recognize hand gestures with high accuracy, surpassing existing studies that have pursued similar goals.
The primary goal of the study is to aid people with hearing or speaking disabilities who struggle with communicating by introducing American Sign Language (ASL) and its corresponding hand gestures. To achieve this goal, the researchers developed a hand gesture detection system using the YOLOv3 algorithm, which can recognize the equivalent alphabet letter of a given gesture. With this system, deaf individuals can conveniently interact and communicate with others.
The study also used tools such as LabelImg for annotating the dataset and categorizing each image of hand gestures based on their corresponding letter alphabet. In the study, Model 18 with a training accuracy of 95.1804%, validation accuracy of 90.8242%, and mAP of 0.8275 was used for the final testing (Alon, Ligayo, Melegrito, Cunanan, & Uy II, 2021). The system was able to detect various hand gestures with over 90% accuracy, making it an effective tool for recognizing sign language.
This project aims to bridge the communication gap between deaf and mute individuals who use Sign Language to communicate with each other and those who do not understand Sign Language. The American Sign Language Lexicon Video dataset was used for the project, and a dominant frame extraction algorithm was employed to create a dataset that matched the requirements. While previous research focused on finger-spellings recognition, this project aimed to identify words as they are more common in communication (Darapaneni, et al., 2021).
The study employed a state-of-the-art Computer Vision model based on Deep Learning to recognize sign language at the pixel level, providing better results by overcoming lighting conditions, orientation, and complexion variables. The study successfully implemented the latest method to train computers to recognize sign languages, creating a communication bridge for those who don't understand Sign Language.
Researchers have been interested in sign alphabet detection for the past decade due to the community's more than 500,000 deaf and mute English-speaking people. While progress has been made in recognizing American Sign Language, both convolution neural networks and traditional machine learning classifiers have been applied in previous studies. This study focused on an American Sign Language dataset with 36 classes of English characters and digits (Hasan, Srizon, Sayeed, & Hasan, 2020). While previous research on this dataset achieved 90% accuracy, this study introduced a modified InceptionV3 architecture for character detection and achieved an overall accuracy of 98.81%, which outperformed all previous studies by a significant margin.
This abstract discusses the challenges of automatic sign language detection as a computer vision problem. Sign language is a non-verbal method of communication used by hard of hearing or deaf individuals, which presents a challenge due to the various modern sign languages and variations in gestures. The study aims to evaluate computer vision-based approaches to sign language recognition and focuses on mapping non-segmented video streams to glosses to gain insights into sign language recognition (Senanayaka, Perera, Rankothge, Usgalhewa, Hettihewa, & Abeygunawardhana, 2022).
The proposed solution is a machine learning model that incorporates Recurrent Neural Network (RNN) layers, including Long Short-Term Memory (LSTM), and is implemented using deep learning frameworks such as Google TensorFlow and Keras API. The goal of the research is to improve the accuracy of sign language recognition and to address the challenges posed by the diversity of sign languages and variations in gestures.
Effective communication skills are essential in everyday life. However, there are times when people struggle to communicate with one another, especially when one party is unable to understand the other. When dealing with those who do not understand sign language, this is a prevalent problem for the deaf-mute community. A model has been designed to assist normal people and enable deaf-mute folks to speak with one another in order to overcome this communication gap. The model is a deep learning-based sign language detection system that identifies American Sign Language (ASL) motions and outputs the associated alphabet in text format (Puchakayala, Nalla, & Pranathi, 2023).
Two models were compared in the study: the CNN model and the YOLOv5 model. The YOLOv5 model was 84.96% accurate, whereas the CNN model was 80.59% accurate. The proposed paradigm aims to enable people with hearing or speaking problems communicate with others more easily and efficiently. This technique bridges the communication gap between the deaf-mute community and others who may not know sign language by identifying ASL motions and converting them into text.
Deaf and hearing individuals alike utilize sign language as a natural and efficient form of communication. ASL or American Sign Language alphabet recognition system along with market-less supported sensors as a difficult endeavor, works for partly challenges to support the variances in signer appearance and in hand segmentation. In this approach, depth information is used instead of colour images to achieve robustness against illumination and background variations.
A simple preprocessing algorithm is applied to the depth image for hand segmentation (Shahin, Aly, & Aly, 2023). Instead of hand-crafted feature extraction methods, feature learning using convolution neural network architectures is applied to the local features extracted from the segmented hand using a simple unsupervised Principal Component Analysis Network (PCANet) deep learning architecture.
This research paper focuses on utilizing deep learning methodologies such as convolution neural networks (CNNs) and computer vision techniques to develop an application that can recognize American Sign Language (ASL) symbols. ASL is a complete language that employs hand movements for communication among people with hearing and speech impairments. The dataset used in this project comprises static hand gesture images captured through a webcam and preprocessed for use in the model building phase (Kiran Babu & Challa, 2022).
The proposed system is based on TensorFlow, Python programming language, and CNN models, and is capable of live detection using a camera feed to provide an accurate and appropriate result of the user's sign. The future scope of this project is vast, as it can be extended to recognize a wide range of hand gestures, words, and even sentences.
Bibliography
Alon, H., Ligayo, M., Melegrito, M., Cunanan, C., & Uy II, E. ( 2021). Deep-hand: a deep inference vision approach of recognizing a hand sign language using american alphabet. In 2021 International Conference on Computational Intelligence and Knowledge Economy (ICCIKE), 373-377.
Arora, S., & Sharma, M. (2021). Deep Learning for Brain Tumor Classification from MRI Images. In 2021 Sixth International Conference on Image Information Processing (ICIIP), 409-412.
Baillarin, S., Tinel , C., Lassalle, P., Melet, O., Youssefi, D., Kettig, P., . . . Gaudissart, V. (2019). Perspectives for VHR Big Data Image Processing and Analytics Toward a Dedicated Framework for Major Disaster and Environment Monitoring from Space. . In IGARSS 2019-2019 IEEE International Geoscience and Remote Sensing Symposium, 5401-5404.
Belhekar, A., Gagare, K., Bedse, R., & Bhelkar, Y. (2019). Leukemia Cancer Detection Using Image Analytics:(Comparative Study). In 2019 5th International Conference On Computing, Communication, Control And Automation (ICCUBEA), 1-6.
Bhadra, R., & Kar, S. (2021). Sign language detection from hand gesture images using deep multi-layered convolution neural network. In 2021 IEEE Second International Conference on Control, Measurement and Instrumentation (CMI), 196-200.
Chen, W., Lei, Y., Liu, M., Zhang, J., & Yao, X. (2020). Workpiece Detection Based on Image Processing and Convolutional Neural Network. In 2020 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD), 302-306.
Darapaneni, N., Gandole, P., Ramasamy, S., Tambe, y., Dwivedi, A., Paduri, A., . . . Ganeshan, K. ( 2021). American sign language detection using instance-based segmentation. In 2021 IEEE International IOT, Electronics and Mechatronics Conference (IEMTRONICS) , 1-6.
Dewantara, D., Hidayat, R., Susanto, H., & Arymurthy, A. (2020). CNN with multi stage image data augmentation methods for Indonesia rare and protected orchids classification. In 2020 International Conference on Computer Science and Its Application in Agriculture (ICOSICA), 1-5.
Du, S., Hong, J., Wang, Y., Xing, K., & Qiu, T. ( 2021). Physical-related feature extraction from simulated SAR image based on the adversarial encoding network for data augmentation. IEEE Geoscience and Remote Sensing Letters, 19, 1-5.
Duong, H., & Hoang, V. (2019). Data augmentation based on color features for limited training texture classification. In 2019 4th International Conference on Information Technology (InCIT) , 208-211.
Hasan, M., Srizon, A., Sayeed, A., & Hasan, M. (2020). Classification of American sign language by applying a transfer learned deep convolutional neural network. In 2020 23rd International Conference on Computer and Information Technology (ICCIT) , 1-6.
Hoque, O., Jubair, M., Islam, M., Akash, A., & Paulson, A. (2018). Real time bangladeshi sign language detection using faster r-cnn. In 2018 international conference on innovation in engineering and technology (ICIET) , 1-6.
Htet, S., Aung, S., & Aye, B. (2022). Real-Time Myanmar Sign Language Recognition Using Deep Learning. In 2022 International Conference on Industrial Engineering, Applications and Manufacturing (ICIEAM), 847-853.
Jindal, N., Yadav, N., Nirvan, N., & Kumar, D. (2022). Sign Language Detection using Convolutional Neural Network (CNN). In 2022 IEEE World Conference on Applied Intelligence and Computing (AIC) , 354-360.
Kim, E., Kim, J., Kim, B., & Kim, S. (2022). Data Augmentation Method based on Feature Extraction for Improving Classification Performance of CNN. In 2022 13th Asian Control Conference (ASCC), 1126-1130.
Kiran Babu, T., & Challa, M. (2022). Recognition of American Sign Language using Deep Learning. Specialusis Ugdymas, 1(43), 4069-4075.
Kumar, T., & Ponnusamy, R. (2022). Automated Chest X-Ray Image Classification using Manta Ray Optimization with Deep Learning Approach. In 2022 International Conference on Augmented Intelligence and Sustainable Systems (ICAISS), 178-184.
Puchakayala, A., Nalla, S., & Pranathi, K. (2023). American Sign language Recognition using Deep Learning. . In 2023 7th International Conference on Computing Methodologies and Communication (ICCMC), 151-155.
Senanayaka, S., Perera, R., Rankothge, W., Usgalhewa, S., Hettihewa, H., & Abeygunawardhana, P. (2022). Continuous American Sign Language Recognition Using Computer Vision And Deep Learning Technologies. In 2022 IEEE Region 10 Symposium (TENSYMP), 1-6.
Shahin, A., Aly, S., & Aly, W. ( 2023). A novel multi-class brain tumor classification method based on unsupervised PCANet features. . Neural Computing and Applications, 1-17.
Shahriar, S., Siddiquee, A., Islam, T., Ghosh, A., Chakraborty, R., Khan, A., & Shahnaz, C. (2018). Real-time american sign language recognition using skin segmentation and image category classification with convolutional neural network and deep learning. In TENCON 2018-2018 IEEE Region 10 Conference (pp. 1168-1171)., 1168-1171.
Susa, J., Macalisang, J., Sevilla, R., Evangelista, R., Quismundo, A., & Reyes, R. (2022). Implementation of Security Access Control using American Sign Language Recognition via Deep Learning Approach. . In 2022 International Conference on Emerging Technologies in Electronics, Computing and Communication (ICETECC), 1-5.
Thakur, K. A. (2021). Manhole Management using Image Processing and Data Analytics. 12th International Conference on Computing Communication and Networking Technologies (ICCCNT), 1-5.
Unkule, P., Shinde, C., Saurkar, P., Agarkar, S., & Verma, U. (2022). CNN based Approach for Sign Recognition in the Indian Sign language. In 2022 International Conference on Augmented Intelligence and Sustainable Systems (ICAISS), 92-97.
Urmee, P., Al Mashud, M., Akter, J. J., & Islam, S. (2019). Real-time bangla sign language detection using xception model with augmented dataset. In 2019 IEEE International WIE Conference on Electrical and Computer Engineering (WIECON-ECE), 1-5.
Wen, Y., & Shi, P. (2007). Image PCA: A new approach for face recognition. . In 2007 IEEE International Conference on Acoustics, Speech and Signal Processing-ICASSP'07, I-1241.
Zhao, Z. ( 2022). Research on single-character image classification of Tibetan ancient books based on deep learning. In 2022 3rd International Conference on Computer Vision, Image and Deep Learning & International Conference on Computer Engineering and Applications (CVIDL & ICCEA), 1-5.
Zulkarnain, I., Nurmalasari, R., & Azizah, F. ( 2022). Table Information Extraction Using Data Augmentation on Deep Learning and Image Processing. In 2022 16th International Conference on Telecommunication Systems, Services, and Applications (TSSA), 1-6.