Misbah Bibi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Misbah Bibi has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- DoHyeun Kim
Department of Computer Engineering, Jeju National University, Jeju, Republic of Korea
9 shared publications - Qazi Waqas Khan
Department of Computer Engineering, Jeju National University, Jejusi 63243, Republic of Korea
7 shared publications - Rashid Ahmad
Faculty of Computing and IT, Sohar University, Sohar, Sultanate of Oman
7 shared publications - Anam Nawaz Khan
Department of Industrial and Systems Engineering, University of Tennessee, Knoxville, TN, 37996, USA.
5 shared publications - Muhammad Faseeh
Department of Electronic Engineering, Jeju National University, Jeju, Republic of Korea
4 shared publications - Atif Rizwan
Department of Electronic Engineering, Kyung Hee University, Yongin, Republic of Korea
3 shared publications
Latest Publications
- Open Access
- 5 August 2026
- Published by Elsevier in Neural Networks
- Vol. 205
DAQSyn: A decentralized adaptive quantization-aware synchronous framework in heterogeneous on-device AI networks
- 5 August 2026
- Published by Elsevier in Neural Networks
- Vol. 205
Open AccessDAQSyn: A decentralized adaptive quantization-aware synchronous framework in heterogeneous on-device AI networks
- Open Access
- 4 August 2026
- Published by Elsevier in Internet of Things
- Vol. 39
LHFL: Logical hierarchical federated learning mechanism in distributed physical AI networks
- 4 August 2026
- Published by Elsevier in Internet of Things
- Vol. 39
Open AccessLHFL: Logical hierarchical federated learning mechanism in distributed physical AI networks
- Open Access
- 25 June 2026
- Published by Wiley in Advanced Intelligent Systems
- Vol. 8 (8)
Enhanced Monocular Depth Estimation Using ResNet Backbone and Multiscale Feature Decoding for Precise Object Distance Measurement
Monocular depth estimation is a key capability for intelligent transportation systems and Internet of Things (IoT)‐enabled cyber‐physical environments, where scalable, low‐cost, and real‐time perception is essential. Traditional approaches rely on stereo vision or LiDAR, increasing deployment cost and limiting large‐scale adoption. This paper presents a deep learning‐based framework that produces high‐fidelity depth maps for long‐range scenes up to 250 m using the D‐Far250 dataset. The architecture employs a ResNet‐based encoder–decoder with multiscale supervision to improve depth consistency across near and far‐range regions. To enable practical perception for autonomous systems, the depth module is integrated with the YOLOv8 object detector to estimate object‐level distances directly from monocular images. Experimental evaluation on both real and synthetic datasets demonstrates strong accuracy and generalization. The model achieves a ( δ 1 ) accuracy of 94.98% on D‐Far250 and 90.85% on the KITTI. The framework operates at over 200 FPS on an NVIDIA RTX 4070 GPU, confirming suitability for latency‐critical IoT applications. By combining accurate monocular depth prediction with detection‐driven distance inference, the proposed system enables scalable, LiDAR‐free perception for collision avoidance, autonomous navigation, unmanned aerial vehicle (UAV) operations, and smart surveillance in connected IoT ecosystems and advanced urban mobility services worldwide and beyond.
- 25 June 2026
- Published by Wiley in Advanced Intelligent Systems
- Vol. 8 (8)
Open AccessEnhanced Monocular Depth Estimation Using ResNet Backbone and Multiscale Feature Decoding for Precise Object Distance Measurement
Monocular depth estimation is a key capability for intelligent transportation systems and Internet of Things (IoT)‐enabled cyber‐physical environments, where scalable, low‐cost, and real‐time perception is essential. Traditional approaches rely on stereo vision or LiDAR, increasing deployment cost and limiting large‐scale adoption. This paper presents a deep learning‐based framework that produces high‐fidelity depth maps for long‐range scenes up to 250 m using the D‐Far250 dataset. The architecture employs a ResNet‐based encoder–decoder with multiscale supervision to improve depth consistency across near and far‐range regions. To enable practical perception for autonomous systems, the depth module is integrated with the YOLOv8 object detector to estimate object‐level distances directly from monocular images. Experimental evaluation on both real and synthetic datasets demonstrates strong accuracy and generalization. The model achieves a ( δ 1 ) accuracy of 94.98% on D‐Far250 and 90.85% on the KITTI. The framework operates at over 200 FPS on an NVIDIA RTX 4070 GPU, confirming suitability for latency‐critical IoT applications. By combining accurate monocular depth prediction with detection‐driven distance inference, the proposed system enables scalable, LiDAR‐free perception for collision avoidance, autonomous navigation, unmanned aerial vehicle (UAV) operations, and smart surveillance in connected IoT ecosystems and advanced urban mobility services worldwide and beyond.
- 7 April 2026
- Published by Elsevier in Information Processing & Management
- Vol. 63 (7)
A distributed quantization mechanism for lightweight federated learning model in heterogeneous on-device AI networks
- 7 April 2026
- Published by Elsevier in Information Processing & Management
- Vol. 63 (7)
A distributed quantization mechanism for lightweight federated learning model in heterogeneous on-device AI networks
- Top Cited PapersOpen Access
- 5 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 6146-6166
Explainable Multi-Modal Fusion-Based Federated Learning for Mortality Prediction in Energy-Constrained Healthcare Systems
The growing demand for privacy-preserving and reliable AI solutions in healthcare has made Federated Learning (FL) a captivating approach for clinical decision support systems. However, contempt its promise, many FL approaches struggle to balance predictive accuracy with explainability, an essential requirement for clinical approval. To address this gap, this study presents an explainable multi-modality FL framework that consolidates a transformer model for unstructured clinical notes with a tree-based model for structured patient data. Using both data modalities, the proposed hybrid architecture empowers robust mortality prediction while ensuring traceability in model decision-making. The framework is evaluated in the large-scale MIMIC-IV dataset within a decentralized hospital environment, where the global fusion-based model achieves an AUC of 0.89 that surpasses traditional baselines while providing interpretable insights on clinical outcomes. These findings validate a scalable and reliable solution for the deployment of explainable AI (XAI) in federated healthcare systems, spanning the critical gap between model performance and clinical confidence.
- 5 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 6146-6166
Top Cited PapersOpen AccessExplainable Multi-Modal Fusion-Based Federated Learning for Mortality Prediction in Energy-Constrained Healthcare Systems
The growing demand for privacy-preserving and reliable AI solutions in healthcare has made Federated Learning (FL) a captivating approach for clinical decision support systems. However, contempt its promise, many FL approaches struggle to balance predictive accuracy with explainability, an essential requirement for clinical approval. To address this gap, this study presents an explainable multi-modality FL framework that consolidates a transformer model for unstructured clinical notes with a tree-based model for structured patient data. Using both data modalities, the proposed hybrid architecture empowers robust mortality prediction while ensuring traceability in model decision-making. The framework is evaluated in the large-scale MIMIC-IV dataset within a decentralized hospital environment, where the global fusion-based model achieves an AUC of 0.89 that surpasses traditional baselines while providing interpretable insights on clinical outcomes. These findings validate a scalable and reliable solution for the deployment of explainable AI (XAI) in federated healthcare systems, spanning the critical gap between model performance and clinical confidence.
- Open Access
- 19 October 2025
- Published by Springer Nature
- p. 111-117
Analyzing Hyperparameter Optimization Methods for Federated Learning Systems
The growing interest in federated learning (FL) has led to increased attention on hyperparameter optimization (HPO) within distributed settings. However, effectively applying HPO in practical FL scenarios presents unique challenges owing to the decentralized characteristics of data and computation. This study provides a comprehensive examination of various HPO strategies tailored for FL, addressing the limitations of earlier analyses that primarily focused on centralized environments with unrestricted data access and consolidated processing resources. In contrast, FL imposes constraints where individual participants operate on isolated local datasets without sharing raw data. As a result, conventional HPO techniques must be adapted to accommodate these decentralized conditions. We explore these adaptations and highlight promising directions for future work. Furthermore, we evaluate the performance of Random Search, a robust yet straightforward HPO approach, through empirical testing in a realistic federated setup using the MNIST dataset in a non-IID setting.
- 19 October 2025
- Published by Springer Nature
- p. 111-117
Open AccessAnalyzing Hyperparameter Optimization Methods for Federated Learning Systems
The growing interest in federated learning (FL) has led to increased attention on hyperparameter optimization (HPO) within distributed settings. However, effectively applying HPO in practical FL scenarios presents unique challenges owing to the decentralized characteristics of data and computation. This study provides a comprehensive examination of various HPO strategies tailored for FL, addressing the limitations of earlier analyses that primarily focused on centralized environments with unrestricted data access and consolidated processing resources. In contrast, FL imposes constraints where individual participants operate on isolated local datasets without sharing raw data. As a result, conventional HPO techniques must be adapted to accommodate these decentralized conditions. We explore these adaptations and highlight promising directions for future work. Furthermore, we evaluate the performance of Random Search, a robust yet straightforward HPO approach, through empirical testing in a realistic federated setup using the MNIST dataset in a non-IID setting.
- 19 October 2025
- Published by Springer Nature
- p. 199-209
Machine Learning-Driven Insights Into Lighting Use in Industrial Buildings
- 19 October 2025
- Published by Springer Nature
- p. 199-209
Machine Learning-Driven Insights Into Lighting Use in Industrial Buildings
- Open Access
- 22 February 2025
- Published by Elsevier in Journal of Building Engineering
- Vol. 103
Dynamic energy cost prediction for industrial and commercial buildings using advanced machine learning and multi-pricing models
- 22 February 2025
- Published by Elsevier in Journal of Building Engineering
- Vol. 103
Open AccessDynamic energy cost prediction for industrial and commercial buildings using advanced machine learning and multi-pricing models
- Open Access
- 6 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 45240-45251
An Auto-Annotation Approach for Object Detection and Depth-Based Distance Estimation in Security and Surveillance Systems
Existing object detection and annotation methods in surveillance systems often suffer from inefficiencies due to manual labeling and a lack of accurate distance estimation, which limits their effectiveness in large-scale environments. These limitations reduce the speed and accuracy required for real-time surveillance, especially in scenarios that necessitate simultaneous monitoring of multiple feeds. To address these challenges, this paper proposes a framework for automated object detection and annotation, specifically designed for surveillance applications. The framework incorporates both manual and automatic modes, offering flexibility in object labeling. A synthetic data is created by using the blender tool which emulates the real-world security and surveillance environments, is utilized to train a yolo model for fast and accurate object recognition and identification. Moreover, for precise distance estimation a depth estimation model is used for calculating the distance for detected objects. The proposed architecture presents both manual and auto modes for the object detection by incorporating the both models in proposed framework. The auto mode of the proposed architecture increases the efficiency of the large surveillance and monitoring applications while lowering the manual labor. The model’s performance is evaluated with state-of-the-art models to assess its performance for auto detection and recognition. The precision and recall above 90% shows that our fine-tuned model demonstrates improved results on synthetic data. With real time surveillance and risk assessment as the foundation, this method seamlessly overthrows drawbacks of existing methods.
- 6 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 45240-45251
Open AccessAn Auto-Annotation Approach for Object Detection and Depth-Based Distance Estimation in Security and Surveillance Systems
Existing object detection and annotation methods in surveillance systems often suffer from inefficiencies due to manual labeling and a lack of accurate distance estimation, which limits their effectiveness in large-scale environments. These limitations reduce the speed and accuracy required for real-time surveillance, especially in scenarios that necessitate simultaneous monitoring of multiple feeds. To address these challenges, this paper proposes a framework for automated object detection and annotation, specifically designed for surveillance applications. The framework incorporates both manual and automatic modes, offering flexibility in object labeling. A synthetic data is created by using the blender tool which emulates the real-world security and surveillance environments, is utilized to train a yolo model for fast and accurate object recognition and identification. Moreover, for precise distance estimation a depth estimation model is used for calculating the distance for detected objects. The proposed architecture presents both manual and auto modes for the object detection by incorporating the both models in proposed framework. The auto mode of the proposed architecture increases the efficiency of the large surveillance and monitoring applications while lowering the manual labor. The model’s performance is evaluated with state-of-the-art models to assess its performance for auto detection and recognition. The precision and recall above 90% shows that our fine-tuned model demonstrates improved results on synthetic data. With real time surveillance and risk assessment as the foundation, this method seamlessly overthrows drawbacks of existing methods.
- Open Access
- 16 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 2679-2693
Real-Time Long-Range Object Tracking Based on Ensembled Model
Accurate Object distance estimation with recognition is essential for various computer vision (CV) applications, including autonomous vehicles and many military operations. While significant advancements have been achieved in supervised and self-supervised techniques for short-range and real-time recognition, existing methods often focus on monocular depth estimation. They are constrained by the limitations of supervised deep learning (DL) models. Taking advantage of temporal information from sequential frames through attention mechanisms to address these challenges offers a promising avenue for enhancing recognition quality. For a real-time military object recognition system, this study integrates RGB images along Depth maps from the KITTI Dataset for short-range measurements. Due to the limitation of the KITTI dataset, a synthetic dataset is generated for long-range object recognition. YOLOv8 is trained for real-time object detection, utilizing the KITTI and a synthetic dataset for long-range analysis. Our method achieves an RMSE of 1.24 meters and an RMSE(log) of 0.18 for depth estimation, outperforming existing approaches. Furthermore, the system accurately detects objects at distances of up to 250 meters, with an average inference time of 15 ms per frame for short-range detection and 18 ms per frame for long-range detection. Comparative evaluations against state-of-the-art methods demonstrate our approach’s superior accuracy and efficiency.
- 16 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 2679-2693
Open AccessReal-Time Long-Range Object Tracking Based on Ensembled Model
Accurate Object distance estimation with recognition is essential for various computer vision (CV) applications, including autonomous vehicles and many military operations. While significant advancements have been achieved in supervised and self-supervised techniques for short-range and real-time recognition, existing methods often focus on monocular depth estimation. They are constrained by the limitations of supervised deep learning (DL) models. Taking advantage of temporal information from sequential frames through attention mechanisms to address these challenges offers a promising avenue for enhancing recognition quality. For a real-time military object recognition system, this study integrates RGB images along Depth maps from the KITTI Dataset for short-range measurements. Due to the limitation of the KITTI dataset, a synthetic dataset is generated for long-range object recognition. YOLOv8 is trained for real-time object detection, utilizing the KITTI and a synthetic dataset for long-range analysis. Our method achieves an RMSE of 1.24 meters and an RMSE(log) of 0.18 for depth estimation, outperforming existing approaches. Furthermore, the system accurately detects objects at distances of up to 250 meters, with an average inference time of 15 ms per frame for short-range detection and 18 ms per frame for long-range detection. Comparative evaluations against state-of-the-art methods demonstrate our approach’s superior accuracy and efficiency.