arXiv subjects

Saad Wazir

Publications and source records attributed to Saad Wazir.

13 recordsLinked to original sources

arXiv2026

HistoSeg++: Delving deeper with attention and multiscale feature fusion for biomarker segmentation

Segmentation of biomarkers in medical images is frequently viewed as a first step towards medical image analysis in any bioinformatics or biomedical application. Despite progress, existing methods still struggle to capture information at multiple scales and to perform upsampling effectively across different datasets. These shortcomings often result in suboptimal generalization capabilities. Recently, architectures belonging to the Nested-UNet family excel in capturing multiscale contextual information and upsample them effectively. In this work, We propose a novel Nested-UNet architecture that effectively captures multi-scale contextual information. It includes inner and outer attention units to enhance focus during upsampling, along with channel-wise feature recalibration using squeeze-and-excitation modules, leading to improved segmentation performance. Additionally, the architecture integrates an edge-aware loss to emphasize boundary accuracy by assigning greater importance to edge regions. Tested extensively on three publicly available benchmark datasets. Our method demonstrates a generalization performance superior to existing Nested-UNet methods. Code: https://github.com/saadwazir/histosegplusplus

cs.CV↗

arXiv2026

MedCAGD: Context-Aware Gated Decoder for Efficient Medical Image Segmentation

Medical image segmentation relies on the ability of encoder-decoder architectures to translate rich feature representations into accurate pixel-level predictions under challenging conditions such as low contrast, structural ambiguity, and scale variability. While recent advances in large-scale pretraining and transformer-based encoders have substantially improved feature extraction, segmentation accuracy remains constrained by decoder design, particularly in terms of cross-scale alignment, contextual integration, and boundary preservation. In this work, we revisit medical image segmentation from a decoder-centric perspective and propose a context-aware gated decoder that systematically regulates feature fusion and contextual aggregation throughout the decoding process. The proposed decoder integrates lightweight multi-scale channel recalibration, gated skip fusion with spatial competition and a global context aggregation mechanism that injects encoder-wide information into intermediate decoding stages. This design enables effective translation of strong pretrained encoder representations into spatially consistent predictions. Extensive experiments across 11 medical image segmentation benchmarks validate the effectiveness and demonstrate that the proposed approach consistently outperforms strong baselines while remaining computationally practical. Code: https://github.com/saadwazir/MedCAGD

cs.CV↗

arXiv2026

TB-AVA: Text as a Semantic Bridge for Audio-Visual Parameter Efficient Finetuning

Audio-visual understanding requires effective alignment between heterogeneous modalities, yet cross-modal correspondence remains challenging when temporally aligned audio and visual signals lack clear semantic correspondence. We propose to use text as a semantic anchor for audio-visual representation learning. To this end, we introduce a parameter-efficient adaptation framework built on frozen audio and visual encoders, centered on Text-Bridged Audio-Visual Adapter (TB-AVA), which enables text-mediated interaction between audio and visual streams. At the core of TB-AVA, Gated Semantic Modulation (GSM) selectively modulates feature channels based on text-inferred semantic relevance. We evaluate the proposed approach on multiple benchmarks, including AVE, AVS, and AVVP, where the proposed framework achieves state-of-the-art performance, demonstrating text as an effective semantic anchor for parameter-efficient fine-tuning (PEFT) in audio-visual learning.

cs.CV↗

arXiv2026

SAT: Selective Aggregation Transformer for Image Super-Resolution

Transformer-based approaches have revolutionized image super-resolution by modeling long-range dependencies. However, the quadratic computational complexity of vanilla self-attention mechanisms poses significant challenges, often leading to compromises between efficiency and global context exploitation. Recent window-based attention methods mitigate this by localizing computations, but they often yield restricted receptive fields. To mitigate these limitations, we propose Selective Aggregation Transformer (SAT). This novel transformer efficiently captures long-range dependencies, leading to an enlarged model receptive field by selectively aggregating key-value matrices (reducing the number of tokens by 97\%) via our Density-driven Token Aggregation algorithm while maintaining the full resolution of the query matrix. This design significantly reduces computational costs, resulting in lower complexity and enabling scalable global interactions without compromising reconstruction fidelity. SAT identifies and represents each cluster with a single aggregation token, utilizing density and isolation metrics to ensure that critical high-frequency details are preserved. Experimental results demonstrate that SAT outperforms the state-of-the-art method PFT by up to 0.22dB, while the total number of FLOPs can be reduced by up to 27\%.

cs.CV↗

arXiv2026

Knowing When to Answer: Adaptive Confidence Refinement for Reliable Audio-Visual Question Answering

We present a formal problem formulation for \textit{Reliable} Audio-Visual Question Answering ($\mathcal{R}$-AVQA), where we prefer abstention over answering incorrectly. While recent AVQA models have high accuracy, their ability to identify when they are likely wrong and their consequent abstention from answering remain underexplored areas of research. To fill this gap, we explore several approaches and then propose Adaptive Confidence Refinement (ACR), a lightweight method to further enhance the performance of $\mathcal{R}$-AVQA. Our key insight is that the Maximum Softmax Probability (MSP) is Bayes-optimal only under strong calibration, a condition usually not met in deep neural networks, particularly in multimodal models. Instead of replacing MSP, our ACR maintains it as a primary confidence signal and applies input-adaptive residual corrections when MSP is deemed unreliable. ACR introduces two learned heads: i) a Residual Risk Head that predicts low-magnitude correctness residuals that MSP does not capture, and ii) a Confidence Gating Head to determine MSP trustworthiness. Our experiments and theoretical analysis show that ACR consistently outperforms existing methods on in- and out-of-disrtibution, and data bias settings across three different AVQA architectures, establishing a solid foundation for $\mathcal{R}$-AVQA task. The code and checkpoints will be available upon acceptance \href{https://github.com/PhuTran1005/R-AVQA}{at here}

cs.LG↗

arXiv2025

Rethinking Decoder Design: Improving Biomarker Segmentation Using Depth-to-Space Restoration and Residual Linear Attention

Segmenting biomarkers in medical images is crucial for various biotech applications. Despite advances, Transformer and CNN based methods often struggle with variations in staining and morphology, limiting feature extraction. In medical image segmentation, where datasets often have limited sample availability, recent state-of-the-art (SOTA) methods achieve higher accuracy by leveraging pre-trained encoders, whereas end-to-end methods tend to underperform. This is due to challenges in effectively transferring rich multiscale features from encoders to decoders, as well as limitations in decoder efficiency. To address these issues, we propose an architecture that captures multi-scale local and global contextual information and a novel decoder design, which effectively integrates features from the encoder, emphasizes important channels and regions, and reconstructs spatial dimensions to enhance segmentation accuracy. Our method, compatible with various encoders, outperforms SOTA methods, as demonstrated by experiments on four datasets and ablation studies. Specifically, our method achieves absolute performance gains of 2.76% on MoNuSeg, 3.12% on DSB, 2.87% on Electron Microscopy, and 4.03% on TNBC datasets compared to existing SOTA methods. Code: https://github.com/saadwazir/MCADS-Decoder

eess.IV↗

arXiv2025

Rethinking the Nested U-Net Approach: Enhancing Biomarker Segmentation with Attention Mechanisms and Multiscale Feature Fusion

Identifying biomarkers in medical images is vital for a wide range of biotech applications. However, recent Transformer and CNN based methods often struggle with variations in morphology and staining, which limits their feature extraction capabilities. In medical image segmentation, where data samples are often limited, state-of-the-art (SOTA) methods improve accuracy by using pre-trained encoders, while end-to-end approaches typically fall short due to difficulties in transferring multiscale features effectively between encoders and decoders. To handle these challenges, we introduce a nested UNet architecture that captures both local and global context through Multiscale Feature Fusion and Attention Mechanisms. This design improves feature integration from encoders, highlights key channels and regions, and restores spatial details to enhance segmentation performance. Our method surpasses SOTA approaches, as evidenced by experiments across four datasets and detailed ablation studies. Code: https://github.com/saadwazir/ReN-UNet

eess.IV↗

arXiv2022

HistoSeg : Quick attention with multi-loss function for multi-structure segmentation in digital histology images

Medical image segmentation assists in computer-aided diagnosis, surgeries, and treatment. Digitize tissue slide images are used to analyze and segment glands, nuclei, and other biomarkers which are further used in computer-aided medical applications. To this end, many researchers developed different neural networks to perform segmentation on histological images, mostly these networks are based on encoder-decoder architecture and also utilize complex attention modules or transformers. However, these networks are less accurate to capture relevant local and global features with accurate boundary detection at multiple scales, therefore, we proposed an Encoder-Decoder Network, Quick Attention Module and a Multi Loss Function (combination of Binary Cross Entropy (BCE) Loss, Focal Loss & Dice Loss). We evaluate the generalization capability of our proposed network on two publicly available datasets for medical image segmentation MoNuSeg and GlaS and outperform the state-of-the-art networks with 1.99% improvement on the MoNuSeg dataset and 7.15% improvement on the GlaS dataset. Implementation Code is available at this link: https://bit.ly/HistoSeg

eess.IV↗

arXiv2021

Classifying Human Activities with Inertial Sensors: A Machine Learning Approach

Human Activity Recognition (HAR) is an ongoing research topic. It has applications in medical support, sports, fitness, social networking, human-computer interfaces, senior care, entertainment, surveillance, and the list goes on. Traditionally, computer vision methods were employed for HAR, which has numerous problems such as secrecy or privacy, the influence of environmental factors, less mobility, higher running costs, occlusion, and so on. A new trend in the use of sensors, especially inertial sensors, has lately emerged. There are several advantages of employing sensor data as an alternative to traditional computer vision algorithms. Many of the limitations of computer vision algorithms have been documented in the literature, including research on Deep Neural Network (DNN) and Machine Learning (ML) approaches for activity categorization utilizing sensor data. We examined and analyzed different Machine Learning and Deep Learning approaches for Human Activity Recognition using inertial sensor data of smartphones. In order to identify which approach is best suited for this application.

cs.HC↗

arXiv2020

Embedded Development Boards for Edge-AI: A Comprehensive Report

The use of Deep Learning and Machine Learning is becoming pervasive day by day which is opening doors to new opportunities in every aspect of technology. Its application Ranges from Health-care to Self-driving Cars, Home Automation to Smart-agriculture, and Industry 4.0. Traditionally the majority of the processing for IoT applications is being done on a central cloud but that has its issues; which include latency, security, bandwidth, and privacy, etc. It is estimated that there will be around 20 Million IoT devices by 2020 which will increase problems with sending data to the cloud and doing the processing there. A new trend of processing the data on the edge of the network is emerging. The idea is to do processing as near the point of data production as possible. Doing processing on the nodes generating the data is called Edge Computing and doing processing on a layer between the cloud and the point of data production is called Fog computing. There are no standard definitions for any of these, hence they are usually used interchangeably. In this paper, we have reviewed the development boards available for running Artificial Intelligence algorithms on the Edge

cs.CY↗

arXiv2020

Single Board Computers (SBC): The Future of Next Generation Pedagogies in Pakistan

ARM processors have taken over the mobile industry from a long time now. Future of data centers and the IT industry is estimated to make use of ARM Processors. Projects like Openstack on ARM are enabling use of ARM in data centers . Single board computers (SBCs) based on ARM processors have become the norm these days. Reason for their popularity lies in their cost effective and power efficient nature. There are hundreds of them available in the market having different sizes, compute power and prices. The reason for their popularity is largely due to the rise of new technology called IoT (Internet of Things) but there is another perspective where they can become handy. Low Price and Power Usage of single board computers makes them top candidate to be used for teaching many courses with hands-on experience in developing countries like Pakistan. Many boards support full Linux distributions and can be used as general-purpose computers while many of them are open hardware based. In this paper, we have reviewed the famous options available and tried to figure out which of them are better for teaching what kind of courses.

cs.CY↗

arXiv2020

HPC as a Service: A naive model

Applications like Big Data, Machine Learning, Deep Learning and even other Engineering and Scientific research requires a lot of computing power; making High-Performance Computing (HPC) an important field. But access to Supercomputers is out of range from the majority. Nowadays Supercomputers are actually clusters of computers usually made-up of commodity hardware. Such clusters are called Beowulf Clusters. The history of which goes back to 1994 when NASA built a Supercomputer by creating a cluster of commodity hardware. In recent times a lot of effort has been done in making HPC Clusters of even single board computers (SBCs). Although the creation of clusters of commodity hardware is possible but is a cumbersome task. Moreover, the maintenance of such systems is also difficult and requires special expertise and time. The concept of cloud is to provide on-demand resources that can be services, platform or even infrastructure and this is done by sharing a big resource pool. Cloud computing has resolved problems like maintenance of hardware and requirement of having expertise in networking etc. An effort is made of bringing concepts from cloud computing to HPC in order to get benefits of cloud. The main target is to create a system which can develop a capability of providing computing power as a service which to further be referred to as Supercomputer as a service. A prototype was made using Raspberry Pi (RPi) 3B and 3B+ Single Board Computers. The reason for using RPi boards was increasing popularity of ARM processors in the field of HPC

cs.DC↗

arXiv2019

Performance Comparison of MPICH and MPI4py on Raspberry Pi-3B Beowulf Cluster

Moore's Law is running out. Instead of making powerful computer by increasing number of transistor now we are moving toward Parallelism. Beowulf cluster means cluster of any Commodity hardware. Our Cluster works exactly similar to current day's supercomputers. The motivation is to create a small sized, cheap device on which students and researchers can get hands on experience. There is a master node, which interacts with user and all other nodes are slave nodes. Load is equally divided among all nodes and they send their results to master. Master combines those results and show the final output to the user. For communication between nodes we have created a network over Ethernet. We are using MPI4py, which a Python based implantation of Message Passing Interface (MPI) and MPICH which also an open source implementation of MPI and allows us to code in C, C++ and Fortran. MPI is a standard for writing codes for such clusters. We have written parallel programs of Monte Carlo's Simulation for finding value of pi and prime number generator in Python and C++ making use of MPI4py and MPICH respectively. We have also written sequential programs applying same algorithms in Python. Then we compared the time it takes to run them on cluster in parallel and in sequential on a computer having 6500 core i7 Intel processor. It is found that making use of parallelism; we were able to outperform an expensive computer which costs much more than our cluster.

cs.DC↗