Thesis & Report, BSc (Computer Science and Engineering)
http://hdl.handle.net/10361/15990
2024-03-28T16:53:16ZEnhancing underwater object detection through water artifact removal and using ensemble transfer learning
http://hdl.handle.net/10361/22189
Enhancing underwater object detection through water artifact removal and using ensemble transfer learning
Saikat, Nayem Hossain; Jahan, Sarowar; Abrar, Fahim; Rahman, Md. Motaqabbir; Rahman, Md. Ashikur
The utilization and exploration of deep-sea resources has made underwater autonomous operation increasingly important to mitigate the dangers of the highpressure deep-sea environment. Intelligent computer vision plays a crucial role in underwater autonomous operation, and pre-processing procedures such as weak illumination and low-quality image enhancement are necessary for underwater vision. Underwater object detection plays a critical role in various domains such as marine biology, environmental monitoring, and underwater robotics. However, it is a challenging task due to the complexities of the underwater environment, including poor visibility, light attenuation, and color distortion. In this research paper, we propose a comprehensive methodology for underwater object detection using transfer learning with PyTorch and Jetson Inference. The contributions of this research paper include advancements in underwater object detection through the combination of transfer learning, fine-tuning, and optimization techniques. The utilization of PyTorch and Jetson Inference frameworks provides a powerful and efficient platform for implementing and deploying the model. Additionally, the incorporation of image-clearing techniques ensures the quality of the dataset and improves the model’s performance in challenging underwater conditions. The results of this research have practical implications for a variety of underwater applications, including marine environment monitoring, underwater exploration, and underwater autonomous robots for visual data collection in complex scenarios. By accurately detecting and classifying underwater objects, our methodology contributes to the understanding and preservation of underwater ecosystems, enhancing the capabilities of underwater systems and facilitating decision-making processes. Future work in this field may involve exploring different architectures, incorporating additional data augmentation techniques, and further fine-tuning the model with larger and more diverse underwater datasets. These efforts will contribute to advancing the state-of-the-art in underwater object detection, enabling more robust and efficient solutions for a wide range of underwater applications. .
This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2023.; Cataloged from PDF version of thesis.; Includes bibliographical references (pages 45-46).
2023-05-01T00:00:00ZOperating and analyzing of business data with digital enterprise division of BanglaCAT
http://hdl.handle.net/10361/22188
Operating and analyzing of business data with digital enterprise division of BanglaCAT
Mitra, Mithila
This internship report provides a comprehensive overview of the experiences, learn- ing, and contributions made during the internship period at Bangla CAT. The re- port encompasses a tailed description of the objectives, scope, methodology, and outcomes of the internship. It also highlights the key projects and tasks under- taken, along with the skills acquired and professional growth achieved during the internship.
This internship report is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2023.; Cataloged from PDF version of thesis.; Includes bibliographical references (page 17).
2023-05-01T00:00:00ZRipple down rule based decision intelligence for mental disorder diagnosis
http://hdl.handle.net/10361/22187
Ripple down rule based decision intelligence for mental disorder diagnosis
Rahman, G.M. Arafat; Ibnul, Tahmid Nizam; Mia, MD. Shamim; Akash, Abid Mahmood; Banerjee, Avinandan
Ripple Down Rule (RDR), a rule-based incremental system, enables knowledge acquisition from human experts to knowledge-based systems (KBS). The majority of modern decision intelligence systems rely on machine learning algorithms, despite the fact that most machine learning algorithms have their own limitations, such as a lack of explainability, an inability to provide multiple outputs, and poor performance with imbalanced or unbalanced data. In addition, RDR still needs to be implemented in the mental health field, and most of the current screening tests cannot diagnose multiple mental disorders at a time. Because of these issues, this paper presents an RDR-based approach for diagnosing mental disorders based on data gathered from primary sources. Since RDR is both a knowledge-based system and an inference engine where domain experts provide rules and conclusions, it can correctly explain its conclusion and provide multiple outputs using the Multiple Classification Ripple Down Rule (MCRDR). In addition, a version of the XGBoost classification algorithm called 'XGBoost Binary Classification Block' has been presented to produce multiple outputs. Comparing the experimental outcomes of three classifier models, we find that XGBoost multi-class classification has 49% accuracy, XGB Binary Classification Block has 96% accuracy, and RDR outperforms the other two by accurately predicting all outputs.
This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2023.; Cataloged from PDF version of thesis.; Includes bibliographical references (pages 59-61).
2023-01-01T00:00:00ZEvaluating the effectiveness of CNN-based models for diabetic retinopathy detection
http://hdl.handle.net/10361/22186
Evaluating the effectiveness of CNN-based models for diabetic retinopathy detection
Niaz, H.M; Tajrian, Nuha; Alam, Mohammad Ahsan Ibn; Limon, Md. Shahriar Khan; Saha, Sharnit
One of the known eye conditions that affect human retinal blood vessels is diabetic
retinopathy (DR). People with diabetes are typically at significantly increased risk
for this. The blood vessels in the retina are damaged when blood sugar levels in the
body increase. Due to the possibility of blindness, people should take precautions
and prioritize early detection. As a result, it is a serious condition because it can
impair vision. It has several stages, including normal, mild, moderate, severe, and
proliferative DR, where it can be quickly determined how severely it has damaged the
retinal blood vessels and the area surrounded by the optical disc. Highly qualified
specialists typically review the colored fundus photos to diagnose this fatal condition. Clinicians struggle to diagnose this condition accurately, and it takes time.
Therefore, several computer vision-based techniques are used to recognize DR and its
various stages from retinal scans automatically. These methods, however, can only
very roughly categorize the various stages of DR because they are unable to capture
the underlying complex properties. However, it is hypothesized that computerized
diagnostic systems using intricate Deep Learning (DL) and convolutional neural network (CNN) structures present a compelling approach to learning about different
patterns of Diabetic Retinopathy (DR) from fundus images, enabling the precise
assessment and categorization of the disease’s severity. This study highlights the
performance summary of CNN-based models EfficientNetV2B3, EfficientNetV2S,
Inception-RestnetV2, MobileNetV2, a fusion model that combines all of these models, and a KNN classifier that uses all of these features that were extracted from each
model to improve the classifications of the stages of DR from these optical fundus
images. This will consequently give the model’s accuracy and a confusion matrix.
In addition, we provide an accurate explanation of the performance of the models
using ExplainableAI. Here, LIME is used for this purpose.
This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2023.; Cataloged from PDF version of thesis.; Includes bibliographical references (pages 51-54).
2023-06-01T00:00:00Z