BRAC University Institutional Repository
Preserving Knowledge, Advancing Research, Sharing Scholarship
A digital platform for collecting, preserving, and sharing BRAC University’s scholarly, academic, and institutional outputs.
Contribute to the Repository
Faculty members and students are invited to submit their research publications, theses, dissertations, and scholarly works to increase visibility, access, and long-term preservation.
Recent Submissions
Faster computation of elemental image generation for real-time integral imaging 3D display system using graphics processing unit and multi-directional projection scheme
(SPIE, 2019-01-01) Alam, Md. Ashraful; Subhani, Mahfuze; Islam, Md. Sifatul; Tareque, Mohd. Zishan; Rafi, M. Rashidur Rahman; Alam M.S.; Kim N.; Department of Computer Science and Engineering
An improved and efficient system for faster computation of Elemental Image generation for real time integral imaging 3D display system with the assistance of Graphics Processing Unit parallel processing is proposed. Previously implemented systems for real time integral imaging system had a resulting frame rate greater than 30 fps for elemental image generation. But this improved and more efficient system is able to produce elemental image at a rate greater than 65 fps for real time integral imaging system. Our proposed model consists of the following steps: information acquisition of objects in real time using Kinect sensor, generation of elemental image sets using pixel mapping algorithm using GPU parallel processing for faster generation. To implement this system, firstly the color (RGB) and depth information data of each object point is acquired from the depth camera (Kinect sensor). Using acquired information, we create the elemental image sets using pixel mapping algorithm. And finally we implemented the pixel mapping algorithm in GPU and hence the overall computational speed of the real time integral display system increased surprisingly. This remarkable increase in speed for elemental image generation opens up new field of possibilities for improvement in integral imaging technology i.e. merging this system with multi-directional projection for real time integral imaging system can enhance the viewing angle remarkably and so on.
Exploring the promise of flavonoids to combat neuropathic pain: From molecular mechanisms to therapeutic implications
(Frontiers Media S.A., 2020-05-01) Uddin M.S.; Al Mamun A.; Rahman M.A.; Kabir, Md. Tanvir; Alkahtani S.; Alanazi I.S.; Perveen A.; Ashraf G.M.; Bin-Jumah M.N.; Abdel-Daim M.M.; Department of Pharmacy
Neuropathic pain (NP) is the result of irregular processing in the central or peripheral nervous system, which is generally caused by neuronal injury. The management of NP represents a great challenge owing to its heterogeneous profile and the significant undesirable side effects of the frequently prescribed psychoactive agents, including benzodiazepines (BDZ). Currently, several established drugs including antidepressants, anticonvulsants, topical lidocaine, and opioids are used to treat NP, but they exert a wide range of adverse effects. To reduce the burden of adverse effects, we need to investigate alternative therapeutics for the management of NP. Flavonoids are the most common secondary metabolites of plants used in folkloric medicine as tranquilizers, and have been claimed to have a selective affinity to the BDZ binding site. Several studies in animal models have reported that flavonoids can reduce NP. In this paper, we emphasize the potentiality of flavonoids for the management of NP.
Glass-free 3D internet TV system using integral imaging
(SPIE, 2019-01-01) Alam, Md. Ashraful; Khan, Amit Hasan; Khan, Fairoz Nower; Nisa, Noor E.; Jannat, Ashraful; Alam M.S.; Kim N.; Department of Computer Science and Engineering
We propose a novel technique to synchronize elemental images and audio signal and the transmission technique for a glass-free 3D TV system based on integral imaging. The main idea behind the method is to generate 3D video based on elemental images synchronized with audio stream. The system uses the depth information and RGB data of per frame of a video through Intel RealSense 3D camera and the audio stream from microphone. The audio file is sampled according to per frame duration of the video and kept in different buffers but having same index. The frames are divided into elemental images using Elemental Image Generation algorithm and the audio signal is synchronized according to the index. Then the stream of elemental images and corresponding audio data is transmitted to data server for storage. HLS streaming protocol is used to stream the TV content. A dedicated web application was made that fetches data from the server and plays video on the user end display device. By using array of micro-lenses in front of display, the video is viewed as three-dimensional with the help of integral imaging technology that omits the need of wearing 3D glasses.
PESCO-BERT: An efficient prompt-based contrastive learning for Bangla news classification
(Institute of Electrical and Electronics Engineers Inc., 2025-01-01) Arman, Mithila; Islam A.; Hoque M.M.; Rahman M.M.
In this work, propose a scalable method for multiclass Bangla news categorization that combines a Bangla-specific data curation pipeline with contrastive, prompt-based fine-tuning of BanglaBERT. Through Unicode and label normalization, punctuation and digit harmonization, source and time-aware splits using shingled-n-gram MinHash, and light minority oversampling, the pipeline hops over label noise, orthographic variation, class imbalance, and data leakage, respectively. Modeling layer. PESCO (Prompt Ensemble Self-Contrastive) and therefore each article is represented as two semantically relevant but stylistically different prompts and trained using a combined loss of weighted cross-entropy and supervised contrastive loss per-class weighting ?=0.5. To make training practical on commodity hardware, employ QLoRA (4-bit NF4 with safe fallbacks), LoRA adapters on attention matrices, gradient checkpointing, mixed precision and conservative micro-batching. BanglaBERT-PESCO achieves 98.89% accuracy and time-aware splits of the Bangla Newspaper Dataset, outperforming other models including BanglaBERT (base and large), XLM-R (base and large), M-BERT, and a QLoRA-tuned LLaMA-3.
Early detection, segmentation and quantification of coronary artery blockage using efficient image processing technique
(SPIE, 2018-01-01) Alam, M.A.; Shakir, M.B.; Hossain M.A.; Pavel M.I.; Shams K.M.A.; Akib F.R.; Department of Computer Science and Engineering
This proposed method aims towards a full automation of the detection of coronary artery blockage through some image processing techniques so that the system does not have to rely on human's inspection. The goal of the research is to implement the proposed image processing techniques so the system can detect the narrowing area of the wall of coronary arteries due to the condensation of different artery blocking agents. The research suggests that the system will require a 64-slice CTA image as input. After the acquisition of the desired input image, it will go through several steps to determine the region of interest. This research proposes a two stage approach that includes the preprocessing stage and decision stage. The pre-processing stage involves common image processing strategies while the decision stage involves the extraction and calculation of two feature ratios to finally determine the intended result. In order to get more insights of the subject of these examinations, this research has proposed the use of an algorithm to create a 3-D model.