Syed Zeeshan Mahmood Shah
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Syed Zeeshan Mahmood Shah has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Farhat Abbas
The Ganj Shakar University (GSU), Pakistan
1 shared publication - Fatima Hassan
Department of Computer Science, COMSATS University of Islamabad, Sahiwal Campus, Pakistan
1 shared publication - Ghulam Ali
Department of Computer Science, University of Okara, Okara, Pakistan
1 shared publication - Mohd Khalid Awang
Universiti sultan zainal abidin
1 shared publication - Mubbashar Saddique
Department of Computer Science & Engineering, University of Engineering & Technology Lahore, Narowal Campus, Narowal, 51601, Pakistan
1 shared publication - Muhammad Tahir Mumtaz
Faculty of Informatics and Computing (FIK), Universiti Sultan Zainal Abidin (UniSZA), Jalan Tembila, 22200 Besut, Terengganu, Malaysia
1 shared publication
Latest Publications
- Open Access
- 14 October 2026
- Published by VFAST Research Platform in VFAST Transactions on Software Engineering
- Vol. 14 (4) , 62-105
Logo Detection and Recognition using Deep Learning: A Systematic Literature Review
Logo detection, recognition, and classification are important computer vision tasks with applications in brand monitoring, content analysis, e-commerce, counterfeit detection, copyright protection, and security systems. Deep learning has advanced logo-analysis systems through automatic feature learning from complex visual data. This paper presents a systematic literature review of deep learning-based logo detection, recognition, classification, localization, matching, and retrieval. The review covers studies published between January 2010 and July 2026. Following database searching, duplicate removal, title-and-abstract screening, full-text eligibility assessment, methodological quality assessment, and final duplicate-publication consolidation, 85 unique studies were included in the final synthesis. The review examines major model families, including convolutional neural networks, region-based CNNs, YOLO variants, SSD, attention-based architectures, and Vision Transformers, discussing their strengths and limitations in detection and recognition performance, computational efficiency, and deployment. It also summarizes evidence on common logo datasets, preprocessing and data preparation strategies, experimental environments, evaluation metrics, and model design choices. Key challenges include limited and imbalanced datasets, background clutter, occlusion, scale and appearance variation, fine-grained brand similarity, cross-domain generalization, open-world recognition, and real-time deployment constraints. Future research directions include lightweight and hardware-aware models, multimodal and open-world learning, few-shot and zero-shot recognition, improved cross-domain robustness, standardized evaluation, and enhanced interpretability. The findings provide a systematic and traceable overview of current progress, methodological limitations, and open research directions in deep learning-based logo analysis.
- 14 October 2026
- Published by VFAST Research Platform in VFAST Transactions on Software Engineering
- Vol. 14 (4) , 62-105
Open AccessLogo Detection and Recognition using Deep Learning: A Systematic Literature Review
Logo detection, recognition, and classification are important computer vision tasks with applications in brand monitoring, content analysis, e-commerce, counterfeit detection, copyright protection, and security systems. Deep learning has advanced logo-analysis systems through automatic feature learning from complex visual data. This paper presents a systematic literature review of deep learning-based logo detection, recognition, classification, localization, matching, and retrieval. The review covers studies published between January 2010 and July 2026. Following database searching, duplicate removal, title-and-abstract screening, full-text eligibility assessment, methodological quality assessment, and final duplicate-publication consolidation, 85 unique studies were included in the final synthesis. The review examines major model families, including convolutional neural networks, region-based CNNs, YOLO variants, SSD, attention-based architectures, and Vision Transformers, discussing their strengths and limitations in detection and recognition performance, computational efficiency, and deployment. It also summarizes evidence on common logo datasets, preprocessing and data preparation strategies, experimental environments, evaluation metrics, and model design choices. Key challenges include limited and imbalanced datasets, background clutter, occlusion, scale and appearance variation, fine-grained brand similarity, cross-domain generalization, open-world recognition, and real-time deployment constraints. Future research directions include lightweight and hardware-aware models, multimodal and open-world learning, few-shot and zero-shot recognition, improved cross-domain robustness, standardized evaluation, and enhanced interpretability. The findings provide a systematic and traceable overview of current progress, methodological limitations, and open research directions in deep learning-based logo analysis.
- Open Access
- 24 February 2026
- Published by ADBA Information Technology and Publishing Limited Company in Computational Systems and Artificial Intelligence
- Vol. 2 (1) , 38-44
Transfer-Learned Convolutional Networks for Multiclass Recognition of Myocardial Infarction and Cardiac Abnormalities in ECG Images
Accurate interpretation of electrocardiograms (ECGs) is central to the early recognition of myocardial infarction and other cardiac abnormalities, yet visual assessment may be affected by subtle inter-class similarities, recording artifacts, and reader-dependent variability. This study presents a controlled comparison of three transfer-learned convolutional neural networks, ResNet-50, ConvNeXt-Tiny, and DenseNet-201, for four-class classification of ECG images. The dataset comprised 3,951 images assigned to abnormal heartbeat, history of myocardial infarction, myocardial infarction, and normal ECG categories. A stratified split allocated 3,159 images to training, 393 to validation, and 399 to independent testing. All models were initialized with ImageNet weights, fine-tuned at an input resolution of 224 × 224 pixels, and assessed using accuracy, macro-averaged precision, recall, F1-score, parameter count, floating-point operations, and inference latency. DenseNet-201 achieved the strongest diagnostic performance, with 99.50% accuracy, 99.48% precision, 99.29% recall, and a 99.37% F1-score, while using 18.10 million trainable parameters. ResNet-50 produced closely comparable accuracy (99.25%) and the lowest latency (0.6949 ms per image), indicating a favorable option for time-sensitive deployment. ConvNeXt-Tiny reached 66.67% accuracy and showed substantially weaker generalization under the common training protocol. The findings indicate that dense feature reuse is particularly effective for preserving subtle ECG morphology, whereas residual learning offers an advantageous accuracy–latency balance. External, patient-level validation remains necessary before clinical use, especially because highly accurate image classifiers may exploit dataset-specific acquisition cues in addition to physiologically meaningful patterns.
- 24 February 2026
- Published by ADBA Information Technology and Publishing Limited Company in Computational Systems and Artificial Intelligence
- Vol. 2 (1) , 38-44
Open AccessTransfer-Learned Convolutional Networks for Multiclass Recognition of Myocardial Infarction and Cardiac Abnormalities in ECG Images
Accurate interpretation of electrocardiograms (ECGs) is central to the early recognition of myocardial infarction and other cardiac abnormalities, yet visual assessment may be affected by subtle inter-class similarities, recording artifacts, and reader-dependent variability. This study presents a controlled comparison of three transfer-learned convolutional neural networks, ResNet-50, ConvNeXt-Tiny, and DenseNet-201, for four-class classification of ECG images. The dataset comprised 3,951 images assigned to abnormal heartbeat, history of myocardial infarction, myocardial infarction, and normal ECG categories. A stratified split allocated 3,159 images to training, 393 to validation, and 399 to independent testing. All models were initialized with ImageNet weights, fine-tuned at an input resolution of 224 × 224 pixels, and assessed using accuracy, macro-averaged precision, recall, F1-score, parameter count, floating-point operations, and inference latency. DenseNet-201 achieved the strongest diagnostic performance, with 99.50% accuracy, 99.48% precision, 99.29% recall, and a 99.37% F1-score, while using 18.10 million trainable parameters. ResNet-50 produced closely comparable accuracy (99.25%) and the lowest latency (0.6949 ms per image), indicating a favorable option for time-sensitive deployment. ConvNeXt-Tiny reached 66.67% accuracy and showed substantially weaker generalization under the common training protocol. The findings indicate that dense feature reuse is particularly effective for preserving subtle ECG morphology, whereas residual learning offers an advantageous accuracy–latency balance. External, patient-level validation remains necessary before clinical use, especially because highly accurate image classifiers may exploit dataset-specific acquisition cues in addition to physiologically meaningful patterns.