Thesis (Bachelor of Science in Computer Science)

Browse

End date: 2019Subject: search.filters.subject.Convolutional Neural Network

Search Results

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    Comparative study of X-ray and CT scan images for the detection of COVID-19 using deep learning
    (BRAC University, 2015-08) Niloy, Ahashan Habib; Shiba, Shammi Akhter; Fahim, S.M. Farah Al; Faria, Faizun Nahar; Rahman, Md. Jamilur; Parvez, Mohammad Zavid
    Coronavirus 2019 (in short, COVID-19), originated in the Wuhan province of China in December 2019, has been declared a global pandemic by WHO in March 2020. Since its inception, it’s rapid spread among nations had initially collapsed the world economy and the increasing death-pool created a strong fear among people as the virus spread through human contact. Initially doctors struggled to diagnose the increasing number of patients as there was less availability of testing kits and failed to treat people efficiently which ultimately led to the collapse of the health sector of several countries. To help doctors primarily diagnose the virus, researchers around the world have come up with some radiology imaging techniques using the Convo lutional Neural Network (CNN). While some of them worked on x-ray images and some others on CT scan images, none worked on both the image types. Thus there’s no way to know which image works better for a particular model. This, therefore, insisted us to perform a comparison between x-ray and CT scan images. Thus we came up with a novel CNN model named CoroPy which works for both the image types and shows that in 2 classes (normal and covid), CT scan images show a better accuracy and it is 99.17% whereas it is 95.73% for x-ray images. However, in the case of 3 classes (normal, covid and viral pneumonia), x-ray images show a better accuracy and it is 92.45% whereas it is 68.81% for CT scan images.
  • Thumbnail Image
    Item
    Human Activity Recognition using wearable body sensor by machine learning approach
    (BRAC University, 2019-12) Promi, Sadia Tangim; Rahman, Md. Zahidur; Mostafa, Moumita; Harun, Sarah Bintay; Alam, Md. Golam Rabiul
    The prevalence of electronics devices and the increase in computer resources, like networking, storage, accessibility and sensor capacity, have significantly improved the lives of humans. Now a days most smart devices have a number of strong sensing equipment, such as sensors for movement, position, connection and direction.Basically, movement or motion tracking sensors are commonly been using to classify the physical activities of humans. This has opened entryways for a wide range of and intriguing applications with regards to a numerous zones, for example, human healthcare well being and transportation, security system. In this point of view, this research gives a complete, best in class audit of the present circumstance of human activity recognition (HAR) approaches with regards to inertial sensors in electronic portable smartphone devices. Our research started by analyzing the principles of human activities and the entire historical events based on electronics deices such a smartphone, which demonstrate the development in this area over the past few years. Our approach concentrates on the introduction of the means of HAR arrangements with regards to sensors. We propose a methodology which incorporates traditional signal processing techniques with deep learning tools to robustly classify activities from wearable body sensor data. Our proposed methodology achieves a validation accuracy of 96.26% in the WISDM Dataset and is able to recognize human activity from wearable body sensor data robustly.
  • Thumbnail Image
    Item
    Leaf classification by feature extraction using CNN
    (BRAC University, 2019-12) Bhuiyan, Md. Mazharul Islam; Nowshin, Jakia; Jaheen, Atkiya; Chakrabarty, Amitabha
    Plants are an integral part of our nature. The identification and classification of plant leaves has always been a matter of interest for the botanists as well as the laymen. Classification of plant leaves will enable us to know the heritage and details of plants at a glance avoiding the duplication of popular names. This recognition system will be beneficial to different sectors of our society including botanic research, medical field, the study of plant taxonomy etc. As leaves carry a lot of information about plant species, extraction of feature is a better way to classify the leaves. In this paper, we have proposed Convolutional Neural Network (CNN) and analyzed plant leaves with different models. We have collected the dataset from Kaggle. By preprocessing the images and extracting the features we have trained our pre-trained model. In our research, we have chosen three models of CNN which are InceptionV3, VGG16 and MobileNet. MobileNet achieved the highest accuracy of 69.47% with a mean absolute error of 30.26, while VGG16 achieved the lowest accuracy of 57.05% with a mean absolute error of 42.95 and 66.13% accuracy for Inception V3.
  • Thumbnail Image
    Item
    Classi cation of magnetic configurations using machine learning algorithms
    (BRAC University, 2019-08) Bokul, Saffat; Abdus Shukur, Samiha Sabrin Md; Ahmed, Saquib; Alam, Md. Ashraful
    Machine learning is used to carry out e cient studies and analyses in the eld of condensed matter physics. We propose comprehensive machine learning approaches that would classify between magnetic structures. We propose models that are trained on data that has been generated on 3D lattices of Heisenberg model using the physical properties of respective magnetic structures. Models are designed based on three types of classi cations, rst classi cation is done between topologically-protected structures, second on non-topologically-protected structures, thirdly on all structures collectively. To achieve this, convolutional neural network (CNN) and support vector machine (SVM) with principle component analysis (PCA) algorithms have been used. We then make a comparative analysis and nd the most optimal solution. The results show that CNN provides the highest accuracy in the classi cation of topological and non-topological magnetic con gurations.
  • Thumbnail Image
    Item
    Analysis on face recognition based on five different viewpoint of face images using MTCNN and FaceNet
    (BRAC University, 2019-08) Al-imran; Shams, Baniamin; Nasim, Faysal Islam; Hossain, Muhammad Iqbal
    Despite signi cant recent achievements in the eld of face recognition, implementing proper face recognition system by training enough data is still a problem because not everyone has enough photos that we can use to train. The objective of this paper is to implement a proper face recognition system which can successfully recognize known and unknown person by feeding only ve phase images for each person into training dataset. In this eld, accuracy and speed of identi cation is the main issue. There are at least two reasons for the importance behind the research of face recognition which has recently received signi cant attention, especially during the past several years. The rst is the wide range of commercial and law enforcement applications, and the second is the availability of feasible technologies after 30 years of research. In this paper, we present a review of the most successful existing method FaceNet for face recognition technology and how we can use it successfully even though we don't have enough data to train and to encourage researchers to embark on this topic. A brief on general information on this topic is also included to compose an overall review. This review is written by investigating past and ongoing studies done by other researchers related to the same subject.