Text recognition technology in landscape images has become a major area of concern, as the assistive technology industry has used it as an important and crucial tool to serve as a basis for developing discoveries and making major advances in innovative assistive technology solu- tions. Assistive technology has been developed for individuals with different types of disabilities through the use of text recognition in landscape images. These solutions enabled persons with disabilities to be active members of society in various fields such as education, employment and others. Recognizing text in landscape photos is a practical but challenging task due to the large differences in backgrounds, textures, fonts, lighting variable, occlusions, variable orientations, and a immense number of non-text objects in nature which has a form similar to textual ele- ments. The Arabic cursive script poses extra challenges that need further investigation. Text extraction technology is based on pattern recognition as text detection, text recognition, and script identification is required. This paper discusses this ICDAR 2017 MLT Arab dataset and OCR methodology that surpassed the challenges that faced the Arabic text environment af- fected by different font sizes, font styles, image resolution, and opacity of text. Our aim in this project, to present a new solution that enables rapid test detection in landscape images, which takes advantage of developments in machine learning and patterns to be a useful reference for researchers and developers.
14/06/2021
The emergence and spread of COVID-19 disease that is announced in December 2019 have impacted individuals’ economy and lives, as it causes respiratory diseases with various symptoms, including coughing and fever. Diagnosing COVID-19 is important to limit the spread of this virus and reduce the number of infections and deaths. The Real-Time Reverse Transcription– Polymerase Chain Reaction RT-PCRis a biology test used to diagnose COVID- 19; this test takes time to produce results. Also has another problem, sometimes the patient of COVID-19 gets a negative result in the RT-PCRtest. Because of these issues, there is a need for an alternative method that is faster and more accurate than the RT-PCR test. So, we are using the radiography method, which becomes necessary in this pandemic. Using x-rays for patients with clear symptoms will helps doctors to predict who are patients more infected and vulnerable to disease than others. An x-ray can display two-dimensional images of patients. We have collected chest x-ray images from five diverse datasets sources considered as large com- pared to previous studies. We used a deep learning technology as an alternative to the RT-PCR test. We suggested two ways to building a model. The first way developed a type of deep learning model by using a Convolutional Neural Network CNN. The model trained on chest x-ray images in the Visual Geometry Group (VGG-16) model. The second way was a hybrid model that is using features extraction from (VGG-16) and using other classification algorithms like support-vector-machines (SVM), Random-Forests (RF), and Extreme-Gradient-Boosting (XGBoost) to classified patients as normal or COVID-19. The result of a (VGG-16) model and a hybrid model that consists of (VGG16+SVM) was the same with 99.82% accuracy and 100% sensitivity (recall). All of the four models we suggested was reliable; the less accuracy of all of them was 98.73%.
15/05/2021
The success of the Data Center of the College of Computer appears in its ability to acquire accurate and timely data and to manage this data for analyzing and good decision-making. The database management system is the only professional structure for this goal to organize, ma- nipulate and retrieve data efficiently. The main objective of this project is to build a rigid and robust integrated database management system for the Data Center of the College to achieve leadership and excellence in providing accurate data and information on students, faculty mem- bers, and staff in the academic, administrative, and research aspects in the College of Computer. In Phase 1 we started in the introduction to our project as we mentioned the problems and advantages of the idea and also the goal of this project. We also set a plan and timetable for the progress of the project, and then we give a literature review. In it, we explained our background on this project, as we added concepts and definitions about it, and also wrote about 11 scientific papers related to it, after that, we talked about the methodology. The most important things we added are Methodology Approach, We talked about the methodology for the progress of the project in an orderly sequence .and the Type of Selected Method, and here we choose the Waterfall methodology to design the database management system. And Software Design, We describe the program design for our system and the blueprints we used in it. And we have concluded with a Summary. In the Implementation phase, we talked about how we implement our project and we men- tioned the development environment of the procedures, devices, and software used to create our database. In the last phase, we gave a conclusion about our project and aspiration for the future work of this project.
15/05/2021
Skin diseases are more common than other diseases. Skin diseases may be caused by a fungal infection, bacteria, or a virus. The texture and color of the skin can change as a result of the disease. Examples of these diseases are chickenpox, impetigo, scabies, infectious erythema, skin warts, and other infectious skin diseases. Skin disorders are long-term and contagious, it can be detected early and with high accuracy before it become a long-term problem. This research builds a system of skin disease detection using the CNN technique and a pre-trained VGG19 model. In addition, the dataset contains 4500 images that were collected from different sources to train the VGG19 model. Data augmentation technique such as zooming, cropping, and rotating was used. After that, the Adamax optimizer, which is most suitable for the proposed methodology, was used to obtain high accuracy and required results. This study achieved a high accuracy of 99% compared to other similar research. It can be concluded that this system is very reliable that can be integrated to smart schools as part of IOT systems.
14/05/2021
The potential of Machine Learning for the healthcare sector is immense. Several Machine Learning techniques have been used to perform predictive analytics in various fields. Predictive analytics in healthcare can help doctors to make decisions about patient’s health and treatment. In addition, the algorithmic approach can be easily used at the scale of a very large number of patients. The proposed project, integrated into this area, aims to propose a solution for diabetes prediction based on Machine Learning techniques to help doctors in early prediction of diabetes and reduce the hospitalizations related to diabetes disease. In this project, we have proposed a diabetes prediction model using Random Forest (RF) classifier that has shown comparatively high performance. To validate our model, we have make experimental evaluation using Diabetes dataset. We compared RF with Naïve Bayes (NB) classifier using the accuracy metrics. The comparative evaluation have shown interesting results. In addition, we have implemented a Graphical User Interface (GUI) for the proposed model to help doctors to take benefit of the model.
15/04/2021
Food waste is a senior global problem, several socio-economic factors have led to increased surplus food quantities in societies. Here comes the role of charitable institutions for food dona- tions, they collect from donors those who have excess food quantities and then distribute food to the needy, My Food Society is considered one of these charities. Nowadays, technology can contribute to reducing food waste. This project aims to solve this problem by developing a web application by MVC pattern for My Food Society that helps preserve the grace by automating the processes of donating food and distributing it, and this will be under the management of the Society. It will also help increase the Society’s workers’ efficiency and speed up their ac- complishment of operations.
15/03/2021
Sign language is a language that uses hand gestures and facial expressions to communicate. Sign language consists of either static or dynamic gestures. Sign language is the main and only communication tool for the deaf and hard of hearing. Therefore, the deaf cannot interact with non-deaf people without a sign language interpreter. Accordingly, sign language recognition automation has become an important application in artificial intelligence and deep learning. Specifically, the recognition of Arabic sign language has been studied using many smart and traditional methods. However, there is still no published study to recognize the Saudi Sign Language based on Saudi sign language dictionary. This research provides a system to rec- ognize dynamic Saudi sign language based on real-time videos to solve this problem. We will construct a dataset in the proposed system and then build a model using the Convolutional long short-term memory (convLSTM) to recognize dynamic signs. Implementing such a system provides a platform for deaf people to interact with the rest of the world without an interpreter to reduce deaf isolation in society.
22/02/2021
Diabetic Retinopathy (DR) is a state that shows up because of harms in the vessels of the retina. It can happen if a person has type one or two diabetics. Additionally, it happens due to a high level of sugar in the blood. In the begging, there is a simple vision problem that may eventually lead to vision loss. It is a common eye disease found in individuals with diabetes. We aim Through the proposed system to diagnose diabetic retinopathy. By examining images of the retina, extracting features.using morphological operations, and classifying these images using Deep Neural Network (CNN) into two categories: either a healthy eye (non-DR images) or an unhealthy eye (DR images).
22/02/2021
Manual text summarization for large quantities of text files is time consuming, costly and re- quires a lot of efforts. Which can be a tiresome and exhausting process. Automatic Text Summarization(ATS) is becoming more and more important lately due to the huge amounts of textual content created on the internet on a daily basis and keeps growing exponentially. ATS can be classified into two main approaches extractive approach, and abstractive approach. The extractive approach selects the most important sentences from the original text and combines them to form a comprehensive summary without changing the words. The abstractive approach from the original text generates new sentences with different words to convey the meaning in a shorter form. This project focus on the extractive approach and proposed model for text summarization using two summary types: Full Summary and Highlight Summary. The model realized as a tool that contained the two summary types.Based on our results, the summaries generated by our model were a short paragraph and few sentences in bullet point form sum- marizing the original text. The proposed model use the standard Document Understanding Conference 2001 dataset to evaluated by using evaluation tool: Recall-Oriented Understudy for Gisting Evaluation (ROUGE ) and compared with a manual summary for both models. Were the evaluation results show that the proposed model attain the highest score on ROUGE-1 recall metric ”0.5027 ” for Full summary, and for Highlight summary gets ROUGE-1 precision metric ”0.7007”.
20/02/2021
Information credibility is a critical issue in social media. People use social media to get information on different topics. Social media is a place that people shared and exchange infor- mation and use social media as a source of information on different topics. However, informa- tion credibility becomes an issue in social media, especially with the Coronavirus Disease 2019 (COVID-19) epidemic. Misinformation has increased, which affects on peoples health, assessing the credibility of the information becomes necessary. In this project, we proposed a supervised machine learning model to assess information credibility in social media using a combination of content-based features and source-based features.
15/02/2021
Education is one of the most important areas of life that affect the development and progress of societies. Education is still in constant development and investment in developing it is con- sidered one of the most successful businesses these days. Technology has contributed from the beginning of its emergence in the development of educa- tion and improve its output, and one of the most prominent contributions made by technology is the developments in the field of information imaging and the provision of visual content; to facilitating the task of transferring information to learners. In the past, people relied on accurate descriptions and simple handwritten drawings to illus- trate information, but it was a weak method that did not perform its full purpose and learners continued to experience problems in the process of knowledge understanding and absorption. Technology has provided many solutions to display information visually using different imaging techniques, incorporating these images in the educational curriculum and display videos during the process of transferring information has largely improved the educating process,most of this visual information is two-dimensional media, which displays only two dimensions and result in a non-comprehensive vision and less information to be shown; this led to the requirement of providing a lot of pictures showing objects’different sides in the educational materials as well as the necessity that teachers must clarify information more in order to absorb the knowledge by the learners, all of this has greatly affected the education outcomes and the time required to teach people. Three-dimensional imaging technology came as an advanced solution that provides people with a more comprehensive view and clarification of things similar to the natural look of the human eye, three-dimensional objects are illustrated either in simple ways such as photos and videos or in a more sophisticated way, Augmented Reality technology has been used as a means of three- dimensional viewing.In Augmented Reality, only through using special glasses or smartphone application, learners will be able to see virtual objects as if it is on the reality surrounding, but to apply this technology to learners, a large number of glasses or smartphones must be provided to every individual, which affects the time required and is considered rather expensive. Hologram Technology is another way to view three-dimensional objects, it is the technology that displays real three-dimensional images that people are able to see directly by the naked eye, these images are called Holograms and are created by a Hologram projectors, a single projector displays a floating three-dimensional object which anyone can see easily. This project studies the implementation of Hologram Technology in education as a tool to dis- play visual information in a three-dimensional way that outperforms other used ways. The implementation of Hologram Technology in this project covers two main types of education, formal and non-formal education, in formal education, the implementation is done on the school environment, and for the non-formal education, an execution model for the implementation of Hologram Technology in Hajj season in holy Mosques is created. this is implemented in the Kingdom of Saudi Arabia society. This project also improves the implementation of Hologram Technology in education through the development of the controlling method of Hologram projector to make it easy to use and to simplify the interaction with the technology.
15/11/2020
People with Locked-In Syndrome (LIS) have neuromuscular disabilities. Due to these lost and other situations, the concept of Brain-Computer Interface (BCI) established which translate brain waves activity to meaningful commands. One of the BCI applications is P300 speller that provides an alternative communication way for people with neuromuscular disabilities. The spelling process begins with presenting a screen of characters with Row-Column (RC) flashes, Single Character (SC) flashes, or Region-Based (RB) flashes, while recording and processing brain signals, and whenever a change in the signals is detected the system will recognize the desired target. P300 speller began with a single layer paradigm, then researchers developed a multi-layer paradigm which recorded higher accuracy. High accuracy increases time-consuming that results in a poor spelling process. As a result, this effort studied the effect of the user’s mother language in the spelling performance. Two healthy participants experienced in our study that was developed using BCI2000. The spelling process has been done with two spellers: Arabic and English, using RC flashes. The results of spelling with user’s mother language, and spelling with foreign language seems to be effected by the length of the word, e.g. ”Love” in Arabic will be spelt faster than ”Love” in English, as it consists of two letters in the Arabic word, whereas it is four letters in the English word. The study concluded that, the length of the word is the main factor that may affect the spelling performance.
15/11/2020