R. S. Naga Kumar Ganugula
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
R. S. Naga Kumar Ganugula has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
M. Venu Gopala Rao
K.L.University, Vadderswaram, India
1 shared publication- Obbu Chandra Sekhar
Department of Electrical Engineering NIT Delhi Delhi India
1 shared publication Rajasekhar Sankuri
Department of Electrical Engineering, National Institute of Technology
1 shared publicationSri Phani Krishna Karri
Department of Electrical Engineering, National Institute of Technology
1 shared publication- Sri Phani Krishna Karri
Department of Electrical Engineering, National Institute of Technology, Tadepalligudem, India
1 shared publication - Suman Machavarapu
Vignan's Lara Institute of Technology and Sciences, Vadlamudi, India
1 shared publication
Latest Publications
- 23 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 33 (10709908) , 2160-2164
Contrastive Dual-Prototypical Learning for Efficient Multi-Label Ophthalmic Diagnosis
Early screening for ophthalmic diseases is a public health imperative for preventing irreversible vision loss. However, real-world fundus datasets often exhibit multi-label comorbidities and long-tailed distributions, inducing diagnostic bias in conventional deep learning models. This letter proposes the Contrastive Dual-Prototypical Learning (CDPL) framework to address these challenges. CDPL utilizes a two-stage curriculum: Self-Supervised Learning (SSL) pre-training to extract invariant features, followed by orthogonality-regularized dual-prototypical fine-tuning that enforces angular separation to decouple overlapping pathological signals. Evaluated on the ODIR dataset using a ResNet-18 backbone, CDPL achieves a 9.7% Macro-F1 improvement over standard benchmarks. Qualitative t-SNE and Grad-CAM analyses confirm the structural refinement of the metric space and superior sensitivity to rare tail-class pathologies. The framework is designed for extensibility to multi-label medical imaging domains characterised by class imbalance, with theoretical generalizability supported by the dataset-agnostic construction of $\mathcal {L}_{\rm{ortho}}$.
- 23 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 33 (10709908) , 2160-2164
Contrastive Dual-Prototypical Learning for Efficient Multi-Label Ophthalmic Diagnosis
Early screening for ophthalmic diseases is a public health imperative for preventing irreversible vision loss. However, real-world fundus datasets often exhibit multi-label comorbidities and long-tailed distributions, inducing diagnostic bias in conventional deep learning models. This letter proposes the Contrastive Dual-Prototypical Learning (CDPL) framework to address these challenges. CDPL utilizes a two-stage curriculum: Self-Supervised Learning (SSL) pre-training to extract invariant features, followed by orthogonality-regularized dual-prototypical fine-tuning that enforces angular separation to decouple overlapping pathological signals. Evaluated on the ODIR dataset using a ResNet-18 backbone, CDPL achieves a 9.7% Macro-F1 improvement over standard benchmarks. Qualitative t-SNE and Grad-CAM analyses confirm the structural refinement of the metric space and superior sensitivity to rare tail-class pathologies. The framework is designed for extensibility to multi-label medical imaging domains characterised by class imbalance, with theoretical generalizability supported by the dataset-agnostic construction of $\mathcal {L}_{\rm{ortho}}$.
- Open Access
- 20 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Multi-Class Classification of Fundus Images through an Enriched-Feature Deep Learning Model for Enhanced Diagnostic Accuracy
Diabetic Retinopathy (DR) and Glaucoma (GL) are critical eye diseases that can lead to vision loss if not identified early. Manual diagnosis using fundus images is time-consuming and error-prone, highlighting the need for automated methods for efficient and accurate classification. Deep learning models improve diagnostic performance but still face challenges such as class imbalance and similarities between disease stages, which complicate early detection.To address these challenges, we propose an Enriched Feature-based Multi-Class (EFMC) deep learning classifier for detecting DR and GL from fundus images. The EFMC classifier operates in three stages: extracting domain-specific features through unsupervised learning, enriching these features with multi-scale filters, and classifying them using multiple CNN classifiers with a hybrid loss function. This approach effectively tackles class similarity and imbalance. Experimental results on benchmark datasets show that the EFMC system outperforms state-of-the-art models, achieving 98.75% accuracy and a 98% F1-score.
- 20 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessMulti-Class Classification of Fundus Images through an Enriched-Feature Deep Learning Model for Enhanced Diagnostic Accuracy
Diabetic Retinopathy (DR) and Glaucoma (GL) are critical eye diseases that can lead to vision loss if not identified early. Manual diagnosis using fundus images is time-consuming and error-prone, highlighting the need for automated methods for efficient and accurate classification. Deep learning models improve diagnostic performance but still face challenges such as class imbalance and similarities between disease stages, which complicate early detection.To address these challenges, we propose an Enriched Feature-based Multi-Class (EFMC) deep learning classifier for detecting DR and GL from fundus images. The EFMC classifier operates in three stages: extracting domain-specific features through unsupervised learning, enriching these features with multi-scale filters, and classifying them using multiple CNN classifiers with a hybrid loss function. This approach effectively tackles class similarity and imbalance. Experimental results on benchmark datasets show that the EFMC system outperforms state-of-the-art models, achieving 98.75% accuracy and a 98% F1-score.
- 1 January 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Load frequency control of three unit interconnected multimachine power system with PI and fuzzy controllers
The load on the power system is always varying with respect to time which results in the variation of frequency, thus leading to load frequency control problem (LFC). The variation in the frequency is highly undesirable and maximum acceptable variation in the frequency is ± 0.5Hz. In this paper load frequency control is done by PI controller, which is a conventional controller. This type of controller is slow and does not allow the controller designer to take into account possible changes in operating conditions and non-linearity's in the generator unit. In order to overcome these drawbacks a new fuzzy load frequency controller is presented to quench the deviations in the frequency and the tie line power due to different load disturbances. The effectiveness of the proposed controller is confirmed using MATLAB/SIMULINK software. The results shows that fuzzy controller provides fast response, very less undershoot and negligible peak overshoots with having small state transfer time to reach the final steady state. In this paper for single area, two area and three area the effectiveness of fuzzy controller is verified.
- 1 January 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Load frequency control of three unit interconnected multimachine power system with PI and fuzzy controllers
The load on the power system is always varying with respect to time which results in the variation of frequency, thus leading to load frequency control problem (LFC). The variation in the frequency is highly undesirable and maximum acceptable variation in the frequency is ± 0.5Hz. In this paper load frequency control is done by PI controller, which is a conventional controller. This type of controller is slow and does not allow the controller designer to take into account possible changes in operating conditions and non-linearity's in the generator unit. In order to overcome these drawbacks a new fuzzy load frequency controller is presented to quench the deviations in the frequency and the tie line power due to different load disturbances. The effectiveness of the proposed controller is confirmed using MATLAB/SIMULINK software. The results shows that fuzzy controller provides fast response, very less undershoot and negligible peak overshoots with having small state transfer time to reach the final steady state. In this paper for single area, two area and three area the effectiveness of fuzzy controller is verified.