Md. Hasanul Kabir

Tanveer Ahmed Belal, G. M. Shahariar, Md. Hasanul Kabir

This paper presents a deep learning-based pipeline for categorizing Bengali toxic comments, in which at first a binary classification model is used to determine whether a comment is toxic or not, and then a multi-label classifier is employed to determine which toxicity type the comment belongs to. For this purpose, we have prepared a manually labeled dataset consisting of 16,073 instances among which 8,488 are Toxic and any toxic comment may correspond to one or more of the six toxic categories - vulgar, hate, religious, threat, troll, and insult simultaneously. Long Short Term Memory (LSTM) with BERT Embedding achieved 89.42% accuracy for the binary classification task while as a multi-label classifier, a combination of Convolutional Neural Network and Bi-directional Long Short Term Memory (CNN-BiLSTM) with attention mechanism achieved 78.92% accuracy and 0.86 as weighted F1-score. To explain the predictions and interpret the word feature importance during classification by the proposed models, we utilized Local Interpretable Model-Agnostic Explanations (LIME) framework. We have made our dataset public and can be accessed at - https://github.com/deepu099cse/Multi-Labeled-Bengali-Toxic-Comments-Classification

A. B. M. Ashikur Rahman, Md. Bakhtiar Hasan, Sabbir Ahmed et al.

Handwritten Digit Recognition (HDR) is one of the most challenging tasks in the domain of Optical Character Recognition (OCR). Irrespective of language, there are some inherent challenges of HDR, which mostly arise due to the variations in writing styles across individuals, writing medium and environment, inability to maintain the same strokes while writing any digit repeatedly, etc. In addition to that, the structural complexities of the digits of a particular language may lead to ambiguous scenarios of HDR. Over the years, researchers have developed numerous offline and online HDR pipelines, where different image processing techniques are combined with traditional Machine Learning (ML)-based and/or Deep Learning (DL)-based architectures. Although evidence of extensive review studies on HDR exists in the literature for languages, such as English, Arabic, Indian, Farsi, Chinese, etc., few surveys on Bengali HDR (BHDR) can be found, which lack a comprehensive analysis of the challenges, the underlying recognition process, and possible future directions. In this paper, the characteristics and inherent ambiguities of Bengali handwritten digits along with a comprehensive insight of two decades of state-of-the-art datasets and approaches towards offline BHDR have been analyzed. Furthermore, several real-life application-specific studies, which involve BHDR, have also been discussed in detail. This paper will also serve as a compendium for researchers interested in the science behind offline BHDR, instigating the exploration of newer avenues of relevant research that may further lead to better offline recognition of Bengali handwritten digits in different application areas.

Mk Bashar, Samia Islam, Kashifa Kawaakib Hussain et al.

Multiple object tracking gained a lot of interest from researchers in recent years, and it has become one of the trending problems in computer vision, especially with the recent advancement of autonomous driving. MOT is one of the critical vision tasks for different issues like occlusion in crowded scenes, similar appearance, small object detection difficulty, ID switching, etc. To tackle these challenges, as researchers tried to utilize the attention mechanism of transformer, interrelation of tracklets with graph convolutional neural network, appearance similarity of objects in different frames with the siamese network, they also tried simple IOU matching based CNN network, motion prediction with LSTM. To take these scattered techniques under an umbrella, we have studied more than a hundred papers published over the last three years and have tried to extract the techniques that are more focused on by researchers in recent times to solve the problems of MOT. We have enlisted numerous applications, possibilities, and how MOT can be related to real life. Our review has tried to show the different perspectives of techniques that researchers used overtimes and give some future direction for the potential researchers. Moreover, we have included popular benchmark datasets and metrics in this review.

Refaat Mohammad Alamgir, Ali Abir Shuvro, Mueeze Al Mushabbir et al.

The task of locating and classifying different types of vehicles has become a vital element in numerous applications of automation and intelligent systems ranging from traffic surveillance to vehicle identification and many more. In recent times, Deep Learning models have been dominating the field of vehicle detection. Yet, Bangladeshi vehicle detection has remained a relatively unexplored area. One of the main goals of vehicle detection is its real-time application, where `You Only Look Once' (YOLO) models have proven to be the most effective architecture. In this work, intending to find the best-suited YOLO architecture for fast and accurate vehicle detection from traffic images in Bangladesh, we have conducted a performance analysis of different variants of the YOLO-based architectures such as YOLOV3, YOLOV5s, and YOLOV5x. The models were trained on a dataset containing 7390 images belonging to 21 types of vehicles comprising samples from the DhakaAI dataset, the Poribohon-BD dataset, and our self-collected images. After thorough quantitative and qualitative analysis, we found the YOLOV5x variant to be the best-suited model, performing better than YOLOv3 and YOLOv5s models respectively by 7 & 4 percent in mAP, and 12 & 8.5 percent in terms of Accuracy.

Akib Mohammed Khan, Alif Ashrafee, Fahim Shahriar Khan et al.

Manually inspecting polyps from a colonoscopy for colorectal cancer or performing a biopsy on skin lesions for skin cancer are time-consuming, laborious, and complex procedures. Automatic medical image segmentation aims to expedite this diagnosis process. However, numerous challenges exist due to significant variations in the appearance and sizes of objects with no distinct boundaries. This paper proposes an attention-based residual Double U-Net architecture (AttResDU-Net) that improves on the existing medical image segmentation networks. Inspired by the Double U-Net, this architecture incorporates attention gates on the skip connections and residual connections in the convolutional blocks. The attention gates allow the model to retain more relevant spatial information by suppressing irrelevant feature representation from the down-sampling path for which the model learns to focus on target regions of varying shapes and sizes. Moreover, the residual connections help to train deeper models by ensuring better gradient flow. We conducted experiments on three datasets: CVC Clinic-DB, ISIC 2018, and the 2018 Data Science Bowl datasets and achieved Dice Coefficient scores of 94.35%, 91.68% and 92.45% respectively. Our results suggest that AttResDU-Net can be facilitated as a reliable method for automatic medical image segmentation in practice.

Minhaz Kamal, Fairuz Shaiara, Chowdhury Mohammad Abdullah et al.

Handwriting Recognition has been a field of great interest in the Artificial Intelligence domain. Due to its broad use cases in real life, research has been conducted widely on it. Prominent work has been done in this field focusing mainly on Latin characters. However, the domain of Arabic handwritten character recognition is still relatively unexplored. The inherent cursive nature of the Arabic characters and variations in writing styles across individuals makes the task even more challenging. We identified some probable reasons behind this and proposed a lightweight Convolutional Neural Network-based architecture for recognizing Arabic characters and digits. The proposed pipeline consists of a total of 18 layers containing four layers each for convolution, pooling, batch normalization, dropout, and finally one Global average pooling and a Dense layer. Furthermore, we thoroughly investigated the different choices of hyperparameters such as the choice of the optimizer, kernel initializer, activation function, etc. Evaluating the proposed architecture on the publicly available 'Arabic Handwritten Character Dataset (AHCD)' and 'Modified Arabic handwritten digits Database (MadBase)' datasets, the proposed model respectively achieved an accuracy of 96.93% and 99.35% which is comparable to the state-of-the-art and makes it a suitable solution for real-life end-level applications.

Md. Bakhtiar Hasan, Tasnim Ahmed, Md. Hasanul Kabir

Biometric authentication using gait has become a promising field due to its unobtrusive nature. Recent approaches in model-based gait recognition techniques utilize spatio-temporal graphs for the elegant extraction of gait features. However, existing methods often rely on multi-scale operators for extracting long-range relationships among joints resulting in biased weighting. In this paper, we present HEATGait, a gait recognition system that improves the existing multi-scale graph convolution by efficient hop-extraction technique to alleviate the issue. Combined with preprocessing and augmentation techniques, we propose a powerful feature extractor that utilizes ResGCN to achieve state-of-the-art performance in model-based gait recognition on the CASIA-B gait dataset.

Fahmida Tasnim Lisa, Md. Zarif Hossain, Sharmin Naj Mou et al.

Semi-supervised learning (SSL) has made significant strides in the field of remote sensing. Finding a large number of labeled datasets for SSL methods is uncommon, and manually labeling datasets is expensive and time-consuming. Furthermore, accurately identifying remote sensing satellite images is more complicated than it is for conventional images. Class-imbalanced datasets are another prevalent phenomenon, and models trained on these become biased towards the majority classes. This becomes a critical issue with an SSL model's subpar performance. We aim to address the issue of labeling unlabeled data and also solve the model bias problem due to imbalanced datasets while achieving better accuracy. To accomplish this, we create "artificial" labels and train a model to have reasonable accuracy. We iteratively redistribute the classes through resampling using a distribution alignment technique. We use a variety of class imbalanced satellite image datasets: EuroSAT, UCM, and WHU-RS19. On UCM balanced dataset, our method outperforms previous methods MSMatch and FixMatch by 1.21% and 0.6%, respectively. For imbalanced EuroSAT, our method outperforms MSMatch and FixMatch by 1.08% and 1%, respectively. Our approach significantly lessens the requirement for labeled data, consistently outperforms alternative approaches, and resolves the issue of model bias caused by class imbalance in datasets.

Md. Atiqur Rahman, Nahian Ibn Asad, Md. Mushfiqul Haque Omi et al.

Automatic Traffic Sign Recognition is paramount in modern transportation systems, motivating several research endeavors to focus on performance improvement by utilizing large-scale datasets. As the appearance of traffic signs varies across countries, curating large-scale datasets is often impractical; and requires efficient models that can produce satisfactory performance using limited data. In this connection, we present 'FUSED-Net', built-upon Faster RCNN for traffic sign detection, enhanced by Unfrozen Parameters, Pseudo-Support Sets, Embedding Normalization, and Domain Adaptation while reducing data requirement. Unlike traditional approaches, we keep all parameters unfrozen during training, enabling FUSED-Net to learn from limited samples. The generation of a Pseudo-Support Set through data augmentation further enhances performance by compensating for the scarcity of target domain data. Additionally, Embedding Normalization is incorporated to reduce intra-class variance, standardizing feature representation. Domain Adaptation, achieved by pre-training on a diverse traffic sign dataset distinct from the target domain, improves model generalization. Evaluating FUSED-Net on the BDTSD dataset, we achieved 2.4x, 2.2x, 1.5x, and 1.3x improvements of mAP in 1-shot, 3-shot, 5-shot, and 10-shot scenarios, respectively compared to the state-of-the-art Few-Shot Object Detection (FSOD) models. Additionally, we outperform state-of-the-art works on the cross-domain FSOD benchmark under several scenarios.

Mehedi Ahamed, Radib Bin Kabir, Tawsif Tashwar Dipto et al.

This study investigates the performance of few-shot learning (FSL) approaches in recognizing Bangla handwritten characters and numerals using limited labeled data. It demonstrates the applicability of these methods to scripts with intricate and complex structures, where dataset scarcity is a common challenge. Given the complexity of Bangla script, we hypothesize that models performing well on these characters can generalize effectively to languages of similar or lower structural complexity. To this end, we introduce SynergiProtoNet, a hybrid network designed to improve the recognition accuracy of handwritten characters and digits. The model integrates advanced clustering techniques with a robust embedding framework to capture fine-grained details and contextual nuances. It leverages multi-level (both high- and low-level) feature extraction within a prototypical learning framework. We rigorously benchmark SynergiProtoNet against several state-of-the-art few-shot learning models: BD-CSPN, Prototypical Network, Relation Network, Matching Network, and SimpleShot, across diverse evaluation settings including Monolingual Intra-Dataset Evaluation, Monolingual Inter-Dataset Evaluation, Cross-Lingual Transfer, and Split Digit Testing. Experimental results show that SynergiProtoNet consistently outperforms existing methods, establishing a new benchmark in few-shot learning for handwritten character and digit recognition. The code is available on GitHub: https://github.com/MehediAhamed/SynergiProtoNet.

ASM Nazrul Islam, Md. Hasanul Kabir, Md. Liakot Ali et al.

Accurate short-term demand forecasting is crucial to airline revenue management, yet most existing systems fail to meet this need because current models treat booking data as a single temporal dimension, either the accumulation of bookings for a specific flight or the historical booking profile of the same route. This unidimensional view discards information carried by the other temporal stream and forecasting absolute passenger counts introduces a further operational fragility when change in planned aircraft type alters total seat capacity. This study addresses both limitations. A dual-stream Long Short-Term Memory (LSTM) integrated with attention framework is proposed that simultaneously processes two complementary input sequences: a horizontal sequence capturing intra-flight booking accumulation over the days preceding departure, and a vertical sequence capturing inter-flight booking patterns at fixed days-before-departure offsets across historical flights. Multiple dual-stream architectural variants, combining self-attention, cross-attention, and hybrid attention with concatenation, residual, and gated fusion strategies, are developed and evaluated. Experiments on real-world reservation data from the national airline of Bangladesh, Biman Bangladesh Airlines (BBA), demonstrate that the proposed hybrid model achieves a Mean Absolute Error of 2.8167 and a coefficient of determination ($R^{2}$) of 0.9495, outperforming single-stream baselines, tree-based models, and three prior dual-LSTM architectures applied to the same data. Validation across four flight category pairs; domestic versus international, direct versus transit, high versus low frequency, and short versus mid versus long haul confirms that the model generalizes across operationally diverse route types. Biman Bangladesh Airlines (BBA) has officially integrated this methodology into its operations.

Sabbir Ahmed, Md. Bakhtiar Hasan, Tasnim Ahmed et al.

To ensure global food security and the overall profit of stakeholders, the importance of correctly detecting and classifying plant diseases is paramount. In this connection, the emergence of deep learning-based image classification has introduced a substantial number of solutions. However, the applicability of these solutions in low-end devices requires fast, accurate, and computationally inexpensive systems. This work proposes a lightweight transfer learning-based approach for detecting diseases from tomato leaves. It utilizes an effective preprocessing method to enhance the leaf images with illumination correction for improved classification. Our system extracts features using a combined model consisting of a pretrained MobileNetV2 architecture and a classifier network for effective prediction. Traditional augmentation approaches are replaced by runtime augmentation to avoid data leakage and address the class imbalance issue. Evaluation on tomato leaf images from the PlantVillage dataset shows that the proposed architecture achieves 99.30% accuracy with a model size of 9.60MB and 4.87M floating-point operations, making it a suitable choice for real-life applications in low-end devices. Our codes and models are available at https://github.com/redwankarimsony/project-tomato.

Md. Safirur Rashid, Sabbir Ahmed, Muhammad Usama Islam et al.

Despite the strong performance of Convolutional Neural Networks (CNNs) in disease classification, their effectiveness often depends on access to large annotated datasets, which is an impractical requirement for emerging or rare conditions such as Monkeypox. To overcome this limitation, we propose a few-shot learning (FSL) framework that employs SimpleShot, a lightweight, non-parametric, inductive classifier, for Monkeypox and pox-like skin disease recognition from limited labeled examples. The proposed pipeline passes the skin lesion images through a frozen, pretrained CNN backbone to obtain feature embeddings, which are then classified via SimpleShot using nearest-centroid comparisons in a normalized embedding space. We systematically benchmark six widely used CNN backbones as feature extractors under consistent experimental settings, enabling fair comparison. Experiments on three publicly available datasets (MSLD v1.0, MSID, and MSLD v2.0) are conducted across 2-way, 4-way, and 6-way tasks with 1-shot, 5-shot, and 10-shot configurations. Among all models, MobileNetV2_100 consistently achieves the highest accuracy. In addition, we present a cross-dataset evaluation for Monkeypox classification, revealing that binary Mpox-vs-Others transfer remains comparatively stable while multi-class performance degrades significantly under domain shift. Together, these results demonstrate the practical utility of combining inductive FSL methods with lightweight CNN backbones and highlight the importance of domain robustness for reliable real-world clinical deployment.

Anika Islam, Tasfia Tahsin, Zaarin Anjum et al.

Accurate and timely identification of plant leaf diseases is essential for resilient and sustainable agriculture, yet most deep learning approaches rely on large annotated datasets and computationally intensive models that are unsuitable for data-scarce and resource-constrained environments. To address these challenges we present a few-shot learning approach within a lightweight yet efficient framework that combines domain-adapted MobileNetV2 and MobileNetV3 models as feature extractors, along with a feature fusion technique to generate robust feature representation. For the classification task, the fused features are passed through a Bi-LSTM classifier enhanced with attention mechanisms to capture sequential dependencies and focus on the most relevant features, thereby achieving optimal classification performance even in complex, real-world environments with noisy or cluttered backgrounds. The proposed framework was evaluated across multiple experimental setups, including both laboratory-controlled and field-captured datasets. On tomato leaf diseases from the PlantVillage dataset, it consistently improved performance across 1 to 15 shot scenarios, reaching 98.23+-0.33% at 15 shot, closely approaching the 99.98% SOTA benchmark achieved by a Transductive LSTM with attention, while remaining lightweight and mobile-friendly. Under real-world conditions using field images from the Dhan Shomadhan dataset, it maintained robust performance, reaching 69.28+-1.49% at 15-shot and demonstrating strong resilience to complex backgrounds. Notably, it also outperformed the previous SOTA accuracy of 96.0% on six diseases from PlantVillage, achieving 99.72% with only 15-shot learning. With a compact model size of approximately 40 MB and inference complexity of approximately 1.12 GFLOPs, this work establishes a scalable, mobile-ready foundation for precise plant disease diagnostics in data-scarce regions.

Muhammad Tausif Ul Islam, Shahir Awlad, Sameen Yeaser Adib et al.

Although existing virtual try-on systems have made significant progress with the advent of diffusion models, the current benchmarks of these models are based on datasets that are dominant in western-style clothing and female models, limiting their ability to generalize culturally diverse clothing styles. In this work, we introduce BD-VITON, a virtual try-on dataset focused on Bangladeshi garments, including saree, panjabi and salwar kameez, covering both male and female categories as well. These garments present unique structural challenges such as complex draping, asymmetric layering, and high deformation complexities which are underrepresented in the original VITON dataset. To establish strong baselines, we retrain and evaluate try-on models, namely StableViton, HR-VITON, and VITON-HD on our dataset. Our experiments demonstrate consistent improvements in terms of both quantitative and qualitative analysis, compared to zero shot inference.

Sabbir Ahmed, Md. Bakhtiar Hasan, Tasnim Ahmed et al.

While deep learning-based architectures have been widely used for correctly detecting and classifying plant diseases, they require large-scale datasets to learn generalized features and achieve state-of-the-art performance. This poses a challenge for such models to obtain satisfactory performance in classifying leaf diseases with limited samples. This work proposes a few-shot learning framework, Domain-adapted Expert Network (DExNet), for plant disease classification that compensates for the lack of sufficient training data by combining observations of a number of expert critics. It starts with extracting the feature embeddings as 'observations' from nine 'critics' that are state-of-the-art pre-trained CNN-based architectures. These critics are 'domain adapted' using a publicly available leaf disease dataset having no overlapping classes with the specific downstream task of interest. The observations are then passed to the 'Feature Fusion Block' and finally to a classifier network consisting of Bi-LSTM layers. The proposed pipeline is evaluated on the 10 classes of tomato leaf images from the PlantVillage dataset, achieving promising accuracies of 89.06%, 92.46%, and 94.07%, respectively, for 5-shot, 10-shot, and 15-shot classification. Furthermore, an accuracy of 98.09+-0.7% has been achieved in 80-shot classification, which is only 1.2% less than state-of-the-art, allowing a 94.5% reduction in the training data requirement. The proposed pipeline also outperforms existing works on leaf disease classification with limited data in both laboratory and real-life conditions in single-domain, mixed-domain, and cross-domain scenarios.

Zahidul Islam, Mohammad Rukonuzzaman, Raiyan Ahmed et al.

Automatically detecting violence from surveillance footage is a subset of activity recognition that deserves special attention because of its wide applicability in unmanned security monitoring systems, internet video filtration, etc. In this work, we propose an efficient two-stream deep learning architecture leveraging Separable Convolutional LSTM (SepConvLSTM) and pre-trained MobileNet where one stream takes in background suppressed frames as inputs and other stream processes difference of adjacent frames. We employed simple and fast input pre-processing techniques that highlight the moving objects in the frames by suppressing non-moving backgrounds and capture the motion in-between frames. As violent actions are mostly characterized by body movements these inputs help produce discriminative features. SepConvLSTM is constructed by replacing convolution operation at each gate of ConvLSTM with a depthwise separable convolution that enables producing robust long-range Spatio-temporal features while using substantially fewer parameters. We experimented with three fusion methods to combine the output feature maps of the two streams. Evaluation of the proposed methods was done on three standard public datasets. Our model outperforms the accuracy on the larger and more challenging RWF-2000 dataset by more than a 2% margin while matching state-of-the-art results on the smaller datasets. Our experiments lead us to conclude, the proposed models are superior in terms of both computational efficiency and detection accuracy.

Shafkat Farabi, Hasibul Himel, Fakhruddin Gazzali et al.

Action quality assessment (AQA) aims at automatically judging human action based on a video of the said action and assigning a performance score to it. The majority of works in the existing literature on AQA divide RGB videos into short clips, transform these clips to higher-level representations using Convolutional 3D (C3D) networks, and aggregate them through averaging. These higher-level representations are used to perform AQA. We find that the current clip level feature aggregation technique of averaging is insufficient to capture the relative importance of clip level features. In this work, we propose a learning-based weighted-averaging technique. Using this technique, better performance can be obtained without sacrificing too much computational resources. We call this technique Weight-Decider(WD). We also experiment with ResNets for learning better representations for action quality assessment. We assess the effects of the depth and input clip size of the convolutional neural network on the quality of action score predictions. We achieve a new state-of-the-art Spearman's rank correlation of 0.9315 (an increase of 0.45%) on the MTL-AQA dataset using a 34 layer (2+1)D ResNet with the capability of processing 32 frame clips, with WD aggregation.

Sayem Mohammad Siam, Jahidul Adnan Sakel, Md. Hasanul Kabir

Human Computer Interaction (HCI) has been redefined in this era. People want to interact with their devices in such a way that has physical significance in the real world, in other words, they want ergonomic input devices. In this paper, we propose a new method of interaction with computing devices having a consumer grade camera, that uses two colored markers (red and green) worn on tips of the fingers to generate desired hand gestures, and for marker detection and tracking we used template matching with kalman filter. We have implemented all the usual system commands, i.e., cursor movement, right click, left click, double click, going forward and backward, zoom in and out through different hand gestures. Our system can easily recognize these gestures and give corresponding system commands. Our system is suitable for both desktop devices and devices where touch screen is not feasible like large screens or projected screens.