Browsing by Author "Islam, Md. Azharul"

Now showing 1 - 10 of 10

Analysis of uncertainty in different neural network structures using monte carlo dropout
(BRAC University, 2023-01) Islam, Md. Farhadul; Zabeen, Sarah; Bin Rahman, Fardin; Islam, Md. Azharul; Bin Kibria, Fahmid; Rasel, Annajiat Alim
Deep learning technologies developed at an exponential rate throughout the years. Starting from Convolutional Neural Networks (CNNs) to Involutional Neural Net works (INNs), there are several neural network (NN) architectures today, including Vision Transformers (ViT), Graph Neural Networks (GNNs), Recurrent Neural Net works (RNNs) etc. However, uncertainty cannot be represented in these architec tures, which poses a significant difficulty for decision-making given that capturing the uncertainties of these state-of-the-art NN structures would aid in making spe cific judgments. Dropout is one method that may be implemented within Deep Learning (DL) networks as a technique to assess uncertainty. Dropout is applied at the inference phase to measure the uncertainty of these neural network models. This approach, commonly known as Monte Carlo Dropout (MCD), works well as a low-complexity estimation to compute uncertainty. MCD is a widely used approach to measure uncertainty in DL models, but majority of the earlier works focus on only a particular application. Furthermore, there are many state-of-the-art (SOTA) NNs that remain unexplored, with regards to that of uncertainty evaluation. There fore an up-to-date roadmap and benchmark is required in this field of study. Our study revolved around a comprehensive analysis of the MCD approach for assessing model uncertainty in neural network models with a variety of datasets. Besides, we include SOTA NNs to explore the untouched models regarding uncertainty. In addition, we demonstrate how the model may perform better with less uncertainty by modifying NN topologies, which also reveals the causes of a model’s uncertainty. Using the results of our experiments and subsequent enhancements, we also discuss the various advantages and costs of using MCD in these NN designs. While working with reliable and robust models we propose two novel architectures, which provide outstanding performances in medical image diagnosis.
Analysis of uncertainty in different neural network structures using Monte Carlo Dropout
(BRAC University, 2023-01) Islam, Md. Farhadul; Zabeen, Sarah; Rahman, Fardin Bin; Islam, Md. Azharul; Kibria, Fahmid Bin; Rasel, Annajiat Alim; Karim, Dewan Ziaul; Manab, Meem Arafat
Deep learning technologies developed at an exponential rate throughout the years. Starting from Convolutional Neural Networks (CNNs) to Involutional Neural Networks (INNs), there are several neural network (NN) architectures today, including Vision Transformers (ViT), Graph Neural Networks (GNNs), Recurrent Neural Networks (RNNs) etc. However, uncertainty cannot be represented in these architectures, which poses a significant difficulty for decision-making given that capturing the uncertainties of these state-of-the-art NN structures would aid in making specific judgments. Dropout is one method that may be implemented within Deep Learning (DL) networks as a technique to assess uncertainty. Dropout is applied at the inference phase to measure the uncertainty of these neural network models. This approach, commonly known as Monte Carlo Dropout (MCD), works well as a low-complexity estimation to compute uncertainty. MCD is a widely used approach to measure uncertainty in DL models, but majority of the earlier works focus on only a particular application. Furthermore, there are many state-of-the-art (SOTA) NNs that remain unexplored, with regards to that of uncertainty evaluation. Therefore an up-to-date roadmap and benchmark is required in this field of study. Our study revolved around a comprehensive analysis of the MCD approach for assessing model uncertainty in neural network models with a variety of datasets. Besides, we include SOTA NNs to explore the untouched models regarding uncertainty. In addition, we demonstrate how the model may perform better with less uncertainty by modifying NN topologies, which also reveals the causes of a model’s uncertainty. Using the results of our experiments and subsequent enhancements, we also discuss the various advantages and costs of using MCD in these NN designs. While working with reliable and robust models we propose two novel architectures, which provide outstanding performances in medical image diagnosis.
Digital marketing activities of Hero Bangladesh
(Daffodil International University, 2019-07-01) Islam, Md. Azharul
This report covers the overall situation of Social Media Marketing platform and its contribution on brand promotion and customer awareness of one of the most prestigious motorbike brand-Hero. The official Facebook pages, Instagram account, YouTube channel of Ammar Hero Bangladesh is directly controlled by the digital and creative team of WAVEMAKER Bangladesh, which is one of the foremost media buying agencies of GroupM and Asiatic 3sixty Bangladesh. My main objective was to learn the insights of digital marketing and understand the impact of social media on the field of digital marketing that is being constantly established nowadays for the long run success of companies. In this report, I have included the organization’s overview and its activities as well as the responsibilities I have been given to handle. I also included various creative contents published by the official page of insight tool every week. Moreover, this report contains discussion on the process of successful media and digital marketing plans executed by WAVEMAKER Bangladesh. I have prepared this business operation model of the company. The second chapter is about my job responsibilities for the organization and the last chapter is about my learning and challenges I have faced and the theories on which I could relate to my work during my internship period at WAVEMAKER Bangladesh.
Fuzzy Modelling for Prediction of Bursting Strength of Knitted Cotton Fabric Using Bleaching Process Variables
(AATCC Journal of Research, 2019-01-01) Haque, Abu Naser Md. Ahsanul; Smriti, Shamima Akter; Farzana, Nawshin; Siddiqa, Fahmida; Islam, Md. Azharul
A fuzzy prediction model has been built based on hydrogen peroxide concentration, bleaching temperature, and time of bleaching as the input variables and knitted cotton fabric bursting strength as the output variable. Fuzzy expert systems can map efficiently in nonlinear domains with minimal experimental data. The model developed in the present study has been validated by new experimental data. The root means square, mean absolute error percentage, and coefficient of determination (R 2) between the predicted and experimental values were found to be 4.89, 0.707, and 0.965 respectively. The results confirm that the model can be applied successfully for the prediction of fabric bursting strength in textile dye houses.
Pedestrian Walking Speed Data Analysis of Dhaka City In Bangladesh
(Daffodil International University, 2021-07-12) Sarkar, Rathindra Nath; Islam, Raisul; Shuvo, Sumit Shom; Siam, Md. Rakibul Islam; Islam, Md. Azharul
Walking is the most proficient and viable method of transportation for short excursions without cost. The Walking pace of walkers is of prime significance in an investigation of capacity, plan furthermore, arrangement of walker offices. This survey focused on the impacts of individual qualities; land utilizes reason, wireless use, carrying things, and development in gatherings. Progressed research on common stream qualities is a critical prerequisite in Bangladesh for improving its street network offices. Around 5 hours of information were gathered by video recording and an absolute number of 3612 people on foot were surveyed for the model improvement. The mean Walking pace of the Bangladeshi person on foot (65.78 m/min) is discovered to be slower when contrasted with US, Europeans and Asian nations however higher than the walker of Saudi Arabia and Indonesia. A further breakdown of the speed information by sexual orientation shows that the female people on foot (31.35 m/min) male (35.36 m/min) partners. The mean Walking pace of the person on foot of Dhaka metropolitan city (32.66 m/min) is most reduced among four significant urban communities of Bangladesh. The greatest and least mean Walking speed is found at the walkway of Dhanmondi 32, Dhaka (28.99 m/min) . It could be expected the presence of a higher number of walkers. Male walkers are quicker than female passersby in all of the three destinations of Dhaka metropolitan urban areas. Kids pedestrians on foot are the slowest walker in the whole metropolitan urban communities. The speed of three sites all middle-aged (32.90 m/min) and younger (37.34 m/min) people on foot are manually higher from mean Walking speed, however, the Walking velocity of three sites for all older pedestrians is (27.09 m/min). Various quantities of gathering size passerby with higher speed in the walkway of Dhanmondi 32, Dhaka. This survey analyzed the elements which impacted strolling rates to give a more noteworthy comprehension of walker developments. The person on foot speed diminishes with expands the gathering group size. Mean strolling pace of walker group size two is higher than other gathering size walkers in all gender and age bunch. Essentially common of Gathering Size three or four than bigger Gathering size. In any case, in certain locales, remarkable circumstances are happened because of a low number of walkers. Female and Older walkers of Gathering Size four. Kids walkers of Gathering Size two or three. The walking velocities of walkers conveying stuff are most certainly not the same as those not conveying stuff. The walking speeds in blended land utilize are just slower than the mean strolling speeds among all land employments. The male person on foot strolling speed is higher than the female walker in each land use.
Prediction of whiteness index of cotton using bleaching process variables by fuzzy inference system
(2018-02-28) Naser, Abu; Haque, Md. Ahsanul; Smriti, Shamima Akter; Hussain, Manwar; Farzana, Nawshin; Siddiqa, Fahmida; Islam, Md. Azharul
A fuzzy prediction model has been built based on hydrogen peroxide concentration, temperature and time of bleaching as the input variables and knitted fabric whiteness index as the output variable. The process parameters affecting the whiteness index of cotton knitted fabrics are very non-linear. Fuzzy inference system is a prospective modeling tool as it can map effectively in nonlinear domain with minimum investigational data. Triangular-shaped membership functions were considered for the variables and total 48 rules were created in this study. It was found that the sole effect of the concentration of hydrogen peroxide on whiteness is pretty low, but is affected by temperature noticeably even in a fixed concentration of hydrogen peroxide. The model proposed in the present study has been verified by additional experimental data set. The root mean square, mean absolute error percentage and coefficient of determination (R 2 ) between the predicted and experimental values were found to be 0.536, 0.798 and 0.959 respectively. The results validate that the model can be applied suitably for the prediction of fabric whiteness index in textile industries.
Present situation of Gazipur Upa-Sahar Ltd. (on basis of real estate sector of Bangladesh)
(BRAC University, 5/22/2016) Islam, Md. Azharul; Ahmed, Riyashad
The real estate sector is the growth center for the development of an economy. Bangladesh, being one of the densely populated nations in the world, has been experiencing severe of houses shortage for its citizens. Although majority of the population is segmented into the middle and the low income groups, still housing for all has been a fallacy in Bangladesh. The gap between demand and supply is still joy wide. Despite inadequate policy preparations, these real estate developers have been successfully making business although the middle and the low income households are still untapped. Gazipur Upa-Sahar Ltd. was established in the year 2010 with REHAB membership, and Gazipur Up-Sahar Ltd. is one of the growing real estate and developers companies in Gazipur.Gazipur Up-Sahar Ltd. provides quality and innovative real estate and developer’s product to the targeted group. As an intern, I got the chance to attach at GUSL for 12 weeks. The internship is a way to relate practical knowledge with the theoretical knowledge. In this report, I discuss and analysis the real estate industry and GUSL separately. Also I identify some internal and external factors that influence this sector and also the organization. The existing marketing strategies are almost the same to all the developers. The main objective of the company is to find out clients and motivate them to buy apartments. This objective accomplished by several developers in several ways. The company has a good future. It is successfully meeting the consumer demand by providing a wide range of real estate and developers products. It is playing an important role in our economy too by providing quality housing product. It also creates huge employment for the people of our country. The company has many scopes to expand its operation, and it has a future plan for expansion However, to pave the way for the organic development of the industry, the problem of long existent inadequate financing availability at lower burden of terms and costs must have to be removed immediately and wider scope has to be created for Non-Resident Bangladeshis. With many other issues, this study attempts to identify the current status of the private housing real estate in Bangladesh, presents deeper insights of the critical factors for increasing its coverage, and thus finally recommends some immediate measures which will be helpful for both the organization and the industry also.
Psychosocial Factors associated with Health Related Quality of Life among University Students
(University of Dhaka, 2019-02-28) Islam, Md. Azharul
Introduction: Health-related quality of life (HRQoL) is individual‘s subjective perception of his or her health condition that covers physical, psychological and social domains. HRQoL has been increasingly used as an indicator of well-being and outcome measure across various studies including randomized control trials (RCTs). Although a large body of research has explored various factors associated with HRQoL for a range of health conditions, very few investigated that for the university populations. The university groups are critical in relation to health due to their distinct psychosocial and physical characteristics. Exploring modifiable factors associated with HRQoL of this group might open a useful avenue for potential health and wellness interventions. The current research was undertaken to explore HRQoL and its associated psychosocial factors among university students. Method: The study followed a cross-sectional survey design to meet its objectives. Participants were the graduate students of a leading public university of Bangladesh. A total of 588 graduate students from randomly selected five faculties (out of 11) responded to a questionnaire survey. HRQoL was measured using revised Indian (Bengali) adapted Short Form of Health version 2 (SF-12 v 2) questionnaire. Permission was obtained from QualityMetric Inc., the copyright holders of SF-12 v 2 (License agreement # QM030016). Psychological distress and self-esteem were captured using Bangla translated 12-items General Health Questionnaire (GHQ-12) and Rosenberg Self-esteem Scale (RSE), respectively. Information on demographic (e.g., sex, age, relationship status, living status), behavioural (e.g., smoking, physical activity, physical illness), and academic (e.g., faculty, CGPA) factors were recorded in aseparated sheet. Ethical approval was obtained from the concerned university ethics committee. Results: The SF-12 v 2 questionnaire generates eight subscales score and two summary scores. The highest score was found in ‗Physical Functioning‘ subscale out of the eight sub-domains. Females were better in ‗Social Functioning‘ (mean: 67.09 vs. 60.10), ‗Role Emotional‘ (mean: 58.76 vs. 54.08) subscales than males. Males were better in ‗Physical Component Summary (PCS)‘ (mean: 44.71 vs. 43.53) than females. ‗General Health‘ and ‗Mental Health‘ were better for those who had no romantic relationship. ‗Social Functioning‘ was better for those who are in a relationship. Break up in a romantic relationship is associated with poorer scores in all dimensions. Residential students pose higher ‗Mental Component Summary (MCS)‘ than their non-residential counterparts (mean: 44.77 vs. 43.09). As for academic orientation, Arts faculty students reported significantly higher MCS followed by Fine Arts, Sciences, and Business studies. Engaging in physical activity is associated with higher PCS (mean: 45.08 vs. 43.61). Likewise, the absence of physical illness was associated with higher PCS (mean: 44.64 vs. 42.93) and MCS (mean: 44.91 vs. 40.11). The smoking cigarette was also associated with poorer mental health (mean: 41.76 vs. 44.74). Both PCS and MCS were significantly and negatively associated with psychological distress but positively with self-esteem. Multivariate analyses revealed male gender, socio-economic status (SES), non-residential status, psychological distress, and self-esteem as significant predictors of PCS [R2=0.133, F (14, 573) =6.30, p<.0001] with self-esteem (β=0.215, p<.0001) and psychological distress (β=-0.131, p<.01) being the two most crucial predictors. Similarly, non-residential status, theabsence of physical illness, psychological distress and self-esteem all emerging as significant predictors for MCS with psychological distress (β=-0.288, p<.01) and self-esteem (β=0.215, p<.0001) being the two most crucial predictors. Discussion and conclusion: The current study highlighted some key areas that might be critical for the HRQoL of the university students of Bangladesh. Attention should be paid to social-emotional aspects of male students while physical well-being for the female students. Necessary psychological support such individual and/or group counselling for the students going through relationship break-up would be useful to cope with the arisen strain and vulnerability. Specific intervention addressing the mental well-being of non-residential as well as students belonging to sciences, fine arts and business faculties is warranted. Promoting physical activity by creating plenty of games and sports would yield as higher physical and mental health of the students. In addition, extracare should be paid to students suffering from any kind of diseases as it has direct consequences on HRQoL. Furthermore, the campaign against smoking should be strengthened across the campus. Finally, adequate mental health support in the form counselling and psychotherapy, mental health workshops, and seminars, for the students in need could improve students‘ overall health status.
Studies on neurotrans mitter-mediating Enzyme-Dopamine-B-hydroxylase and other Biochemical parameters in Diabetic-and Heart Disease patients
(© University of Dhaka, 2025-04-23) Islam, Md. Azharul
Vehicle Detection Using Deep Learning Techniques
(©Daffodil International University, 2021-06-20) Islam, Md. Azharul
Vehicle detection and classification using deep learning methods has been found out in this paper. In the area of highway management, vehicle detection and classification are becoming more significant currently. Vehicle Detection and Classification based on Multiple Deep Learning Methods has been found in this paper, multiple classes and multiple methods have been used on this topic in very less research paper. In fact, there are different types of vehicles, such as cars, microbuses, jeeps, pickups, buses, trucks, taxis, vans, rickshaws, etc. Multiple vehicles have different shapes and sizes (bounding boxes) so it is very difficult to detect this multiple class, in this paper multiple classes of vehicle have been used. We have used three of the deep learning methods in this paper, method performance, detection ability and object classification has been compared with those methods. The three deep learning methods we have proposed are Mask R-CNN, Faster R-CNN and Yolo V5 method. Here ResNet50 is used as backbone in Faster RCNN method and ResNet101 is used as backbone in Mask R-CNN method, where Mask R-CNN and Faster R-CNN methods are included in CNN family ties. Though the Mask R-CNN is the extension of Faster R-CNN. We evaluate our models' performance through Confusion Matrix. The methods of F1 score, mean average recall and mean average precision have been found out through the Confusion Matrix, the methods have been compared with those values. From that value it is evident that Mask R-CNN gives better performance than other methods. We see from the table (table: 6) that the following values are obtained using Confusion Matrix from Mask R-CNN method F1 score - 87%, mean average recall- 92% and mean average precision - 82%. So The Mask R-CNN's detection score is higher than other models, so the Mask R-CNN's detection ability and classification is better than other models. There will be a lot of cooperation in vehicle detection and prediction for self-driving cars or various robotic cars through this work.