Browsing by Author "Aronya, Biraj Saha"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    A Transfer Learning Approach for Face Recognition Using Average Pooling and MobileNetV2
    (Daffodil International University, 2022-07-01) Shamrat, F. M. Javed Mehedi; Chakraborty, Sovon; Moharram, Md. Shakil; Roy, Tonmoy; Rahman, Masudur; Aronya, Biraj Saha
    Facial recognition is a fundamental method in facial-related science such as face detection, authentication, monitoring, and a crucial phase in computer vision and pattern recognition. Face recognition technology aids in crime prevention by storing the captured image in a database, which can then be used in various ways, including identifying a person. With just a few faces in the frame, most facial recognition systems function sufficiently when the techniques have been tested under artificial illumination, with accurate facial poses and non-blurry images. In our proposed system, a face recognition system is proposed using average pooling and MobileNetV2. The classifiers are implemented after a set of preprocessing steps on the retrieved image data. To compare the model is more effective, a performance test on the result is performed. It is observed from the study that MobileNetV2 triumphs over average pooling with an accuracy rate of 98.89% and 99.01% on training and test data, respectively.
  • No Thumbnail Available
    Item
    Classification of Breast Cancer Cell Images using Multiple Convolution Neural Network Architectures
    (Scopus, 2021) Tasnim, Zarrin; Shamrat, F. M. Javed Mehedi; Islam, Md Saidul; Rahman, Md.Tareq; Aronya, Biraj Saha; Muna, Jannatun Naeem; Billah, Md. Masum
    Abstract: Breast cancer is a malignant tumor that affects women. It is the most prevalent cancer in women, affecting about 10% of all women at any point in their lives. The development of breast cancer begins in the lobules or ducts of the cells. Early detection and prevention are the best ways to stop this cancer from spreading. In this study, five Convolution Neural Network (CNN) models are used to process image data of breast cells. Alex Net, InceptionV3, GoogLeNet, VGG19 and Exception models are used for the classification of Invasive Ductal Carcinoma, IDC and Non-Invasive Ductal Carcinoma (Non-IDC) cells. The models are trained and tested at different epochs to record the learning rate. It is observed from the study that with higher epochs, the data loss decreases and accuracy increases. The accuracy of InceptionV3 and Exception is 92.48% and 90.72% respectively. Likewise, VGG19 and Alex Net have fairly close accuracy of 94.83% and 96.74%. However, GoogLeNet dominates over the other implemented models with the highest accuracy of 97.80%. The GoogLeNet model performs with high accuracy and precision in detecting IDC cells responsible for breast cancer.