Browsing by Author "Md. Shahriar Hussain"

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    Open Access
    Automated Fisheries Water Quality Monitoring System
    (North South University, 2020-04-30) Tanbin Akter Mitaly; Paplu Dash; Md. Shahriar Hussain; 1712438642; 1611420043
    Fisheries is one of the main prospects to grow economically also we benefit from it by the production of food and the nutrition that is required us to live by. For fisheries, it's important to maintain good water quality as it is an important factor that affects fish health and performance in the aquaculture production system. It matters what type of fish we are culturing in our system due to different fish species need a different range of water quality aspects in which they can survive. Different fish species have a different optimized point to growing well. Don't maintain the water quality can result in parasite infestations, poor growth, erratic behaviour and various disease system. The water quality is dependent on climate, seasonal changes also how a culture system is used. To maintain good water quality conditions we need to look at physical and chemical components of water that have an impact on water quality.
  • Thumbnail Image
    Item
    Open Access
    Lip Reading App
    (North South University, 2023) Jubair Mahmud Pulock; Md. Shahriar Hussain; 1931111042
    Effective communication is essential for promoting inclusivity and closing the gap between people who speak different languages in today's globally connected world. But hearing-impaired people frequently have trouble understanding spoken language and must rely significantly on visual signals to follow discussions. To solve this problem, we introduce a revolutionary Lip Reading App, a state-of-the-art mobile application that uses computer vision and artificial intelligence (AI) to enable real-time speech interpretation. In order to effectively monitor and interpret lip movements and translate them into understandable text or audio output, the Lip Reading App makes use of the sophisticated capabilities of deep learning algorithms. The software recognizes and tracks facial landmarks using facial recognition technology, capturing the delicate movements and gestures involved in speech creation. The software can produce accurate transcriptions or audio renderings of the spoken content by processing these acquired visual cues and comparing them to a huge library of phonetic and linguistic patterns. This project presents the development and training of a character-level sequence-to sequence language model using recurrent neural networks (RNNs). The primary objective of the model is to generate text predictions given input sequences of characters. The model employs the 'binary_crossentropy' loss function, although it might be more suitable to use categorical cross-entropy for multi-class classification tasks. Training is performed over multiple epochs, with a learning rate of 1.0e-4 (0.0001) and an average epoch time of 35026 seconds. The model's progress is evaluated based on the loss value, which steadily decreases with each epoch. By the end of 100 epochs, the average loss stands at 63.765. Despite the potential improvements in the loss function selection, the model exhibits promising capabilities in generating coherent and meaningful text predictions. Further fine-tuning and optimization could potentially enhance the performance and versatility of the language model for various text generation tasks.
  • Thumbnail Image
    Item
    Open Access
    Object Detecting Robot Identifying & Picking Up Objects
    (North South University, 2023) Raisa Akhtar; Zarin Musharrat; Alifa Khan; Md. Shahriar Hussain; 2012010042; 1931826642; 1931829642
    This research focuses on the development of an accurate object detection algorithm to aid robots in identifying and picking up objects in cluttered environments. Object detection is crucial for robotic control systems, particularly in industrial automation. Traditional object detection algorithms struggle with objects lacking texture, leading to the adoption of deep learning methods like YOLO (You Only Look Once) for improved performance. The study aims to enhance object localization accuracy, especially in cluttered scenes, to facilitate robot grasping tasks. While basic YOLO models can detect objects, precise localization remains challenging. The research proposes a solution using deep YOLOs for semantic segmentation, which significantly improves object localization accuracy. Despite being time-consuming, YOLOv8 proves highly efficient, achieving an accuracy of 84.7%. The ultimate goal is to implement a robot capable of detecting and identifying five objects (ball, bottle, car, cup, spoon), and then picking and placing them according to user instructions. For the hardware part, we have used the Arduino UNO microcontroller, which is quite budget-friendly. As this is just a small test project, only a robotic arm gripper is used here, so that it can only grab it if it recognizes the object.
  • Thumbnail Image
    Item
    Open Access
    Parkinson’s disease onset detection Using Machine Learning.
    (North South University, 2022) MD Samiul Hasan; Kazal Roy Sagar; Anika Ferdausy; Md. Shahriar Hussain; 1512466042; 1430784642; 1512232042
    Accurately identifying Parkinson’s disease (PD) at an early stage is unquestionably critical for halting its progression and giving patients access to disease-modifying medication. To achieve that goal, the premotor stage of Parkinson’s disease should be closely studied. Based on premotor traits, a unique deep-learning technique is introduced to determine if an individual has Parkinson’s disease or not. In order to detect Parkinson’s disease at an early stage, numerous indicators were used in this study, including Rapid Eye Movement and olfactory loss, cerebrospinal fluid data, and dopaminergic imaging markers. A comparison of the proposed deep learning model with twelve machine learning and ensemble learning methods based on relatively small data sets of 183 healthy individuals and 401 early Parkinson’s disease patients reveals that the designed model achieves the highest accuracy, 96.45 percent on average. We present the feature importance of the PD detection procedure based on the Boosting approach in addition to identifying the PD. Traditional diagnostic procedures, on the other hand, may suffer from subjectivity because they rely on the interpretation of motions that are sometimes subtle to human sight and hence difficult to define, potentially leading to misdiagnosis. Meanwhile, early non-motor symptoms of Parkinson’s disease may be moderate and caused by a variety of other illnesses. As a result, these symptoms are frequently missed, making early PD diagnosis difficult. To overcome these challenges and improve PD diagnosis and assessment procedures, machine learning algorithms for the viii classification of PD and healthy controls or patients with similar clinical presentations have been used. A total of 209 papers were included, with pertinent material retrieved and provided in this review, with an examination of their aims, data sources, data kinds, machine learning methods, and associated outcomes. These results show that machine learning methodologies and novel biomarkers have a strong potential for adaption in clinical decision-making, leading to more systematic, informed PD diagnosis.