A5020766586- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Fuzzy Logic and Control Systems
- Anomaly Detection Techniques and Applications
- Remote-Sensing Image Classification
- Imbalanced Data Classification Techniques
- Industrial Vision Systems and Defect Detection
- Advanced Image Fusion Techniques
- Neural Networks and Applications
- Human Pose and Action Recognition
- Image and Signal Denoising Methods
- Fuzzy Systems and Optimization
- Multimodal Machine Learning Applications
- Machine Learning and Data Classification
- Energy Load and Power Forecasting
- Digital Media Forensic Detection
- Video Surveillance and Tracking Methods
- AI in cancer detection
- Hydrological Forecasting Using AI
- Stock Market Forecasting Methods
- Radiomics and Machine Learning in Medical Imaging
- Flood Risk Assessment and Management
- Vehicle License Plate Recognition
- Hydrology and Watershed Management Studies
- Image Enhancement Techniques
Emory University
2024-2025
The Wallace H. Coulter Department of Biomedical Engineering
2025
Georgia Institute of Technology
2025
Nanyang Technological University
2023-2024
UNSW Sydney
2020-2023
University of Canberra
2020-2023
UNSW Canberra
2020-2022
Indian Institute of Technology Kharagpur
2014-2017
Enabling effective learning using only a few presented examples is crucial but difficult computer vision objective. Few-shot have been proposed to address the challenges, and more recently variational inference-based approaches are incorporated enhance few-shot classification performances. However, current dominant strategy utilized Kullback-Leibler (KL) divergences find log marginal likelihood of target class distribution, while neglecting possibility other probabilistic comparative...
Virtual staining is an artificial intelligence-based approach that transforms pathology images between stain types, such as hematoxylin and eosin (H&E) to immunohistochemistry (IHC), providing a tissue-preserving efficient alternative traditional IHC staining. However, existing methods for translating H&E virtual often fail generate of sufficient quality accurately delineating cell nuclei IHC+ regions. To address these limitations, we introduce VISTA, platform designed translate into IHC. We...
The fusion of LiDAR and camera sensors has demonstrated significant effectiveness in achieving accurate detection for short-range tasks autonomous driving. However, this approach could face challenges when dealing with long-range scenarios due to disparity between sparsity high-resolution data. Moreover, sensor corruption introduces complexities that affect the ability maintain robustness, despite growing adoption domain. We present SaViD, a novel framework comprised three-stage alignment...
Incorporating deep learning (DL) classification models into unmanned aerial vehicles (UAVs) can significantly augment search-and-rescue operations and disaster management efforts. In such critical situations, the UAV's ability to promptly comprehend crisis optimally utilize its limited power processing resources narrow down search areas is crucial. Therefore, developing an efficient lightweight method for scene of utmost importance. However, current approaches tend prioritize accuracy on...
While evolving neuro-fuzzy systems have shown promise for learning from non-stationary streaming data with concept drift, most existing models lack transparency due to the limited interpretability of Takagi-Sugeno fuzzy architecture's linear rule consequents. The limits reliability crucial applications. To address this limitation, paper proposes a new system called X-Fuzz that enhances by integrating LIME technique provide local explanations and evaluates them using faithfulness monotonicity...
This paper presents a novel multi-fake evolutionary generative adversarial network(MFEGAN) for handling imbalance hyperspectral image classification. It is an end-to-end approach in which different objective losses are considered the generator network to improve classification performance of discriminator network. Thus, same has been used as standard classifier by embedding on top discriminating function. The effectiveness proposed method validated through two spatial-spectral data sets. and...
Many real-world classification problems have imbalanced frequency of class labels; a well-known issue known as the "class imbalance" problem. Classic algorithms tend to be biased towards majority class, leaving classifier vulnerable misclassification minority class. While literature is rich with methods fix this problem, dimensionality problem increases, many these do not scale-up and cost running them become prohibitive. In paper, we present an end-to-end deep generative classifier. We...
Seismic inversion is crucial in hydrocarbon exploration, particularly for detecting hydrocarbons thin layers. However, the detection of sparse layers within seismic datasets presents a significant challenge due to ill-posed nature and poor non-linearity problem. While data-driven deep learning algorithms have shown promise, effectively addressing sparsity remains critical area improvement. To overcome this limitation, we propose OrthoSeisnet, novel technique that integrates multi-scale...
A novel objective function based clustering algorithm has been introduced by considering linear functional relation between input-output data and geometrical shape of input data. Noisy points are counted as a separate class remaining good in the set considered clusters. This noise concept taken into proposed to obtain fuzzy partition matrix product space Block orthogonal matching pursuit is applied determine optimal number rules from over specified (clusters). The obtained used premise...
The main objective of this study is to propose an enhanced wind power forecasting (EWPF) transformer model for handling grid operations and boosting market competition. It helps reliable large-scale integration relies in large part on accurate (WPF). proposed evaluated single-step multi-step WPF, compared with gated recurrent unit (GRU) long short-term memory (LSTM) models a dataset. results the indicate that EWPF outperforms conventional neural network (RNN) terms time-series accuracy. In...
Since their introduction in the last few years, conditional generative models have seen remarkable achievements. However, they often need use of large amounts labelled information. By using unsupervised generation conjunction with a clustering inference network, ClusterGAN has recently been able to achieve impressive results. real distribution data is ignored, network can only inferior performance by considering uniform prior based samples. true not necessarily balanced. Consequently, fails...
A Fuzzy C Regression Model (FCRM) distance metric has been used in Competitive Agglomeration (CA) algorithm to obtain optimal number rules or construct fuzzy subspaces whole input output space. To partition matrix data space, a new objective function proposed that can handle geometrical shape of distribution and linear functional relationship between feature space variable. Premise consequence parameters Takagi-Sugeno (TS) model are also obtained from the function. Linear coefficients part...
Traditional oversampling methods are generally employed to handle class imbalance in datasets. This approach is independent of the classifier; thus, it does not offer an end-to-end solution. To overcome this, we propose a three-player adversarial game-based method, where domain-constraints mixture generators, discriminator, and multi-class classifier used. Rather than minority oversampling, (AO) data-space (DO) approach. In AO, generator updates by fooling both however, DO, favoring...
This paper introduces an interval type-2 modified fuzzy c-regression model (IT2MFCRM) clustering algorithm for identifying the structure in TS Fuzzy Model (TSFM). A scaling factor has been used parameters of set. Once, type-1 MFCRM performed to obtain premise membership function, then parameters. Once function is obtained, type reduction technique coefficients consequence Orthogonal Least Square (OLS) method applied determining Finally, IT2MFCRM based validated on two benchmark examples.
This paper presents an iterative Takagi Sugeno Fuzzy Model (TSFM) identification. Interval Type-2 Recursive C-Means (IT2RFCM) clustering algorithm has been used to classify the data space obtain premise variable parameters and Weighted Least Square (WRLS) technique determine consequence of each linear model. IT2RFCM obtained from type-1 by introducing fuzziness parameters. The effectiveness proposed validated on Mackey-Glass time series data.
A rainfall-runoff model predicts surface runoff either using a physically-based approach or systems-based approach. Takagi-Sugeno (TS) Fuzzy models are approaches and popular modeling choice for hydrologists in recent decades due to several advantages improved accuracy prediction over other existing models. In this paper, we propose new developed Gustafson-Kessel (GK) clustering-based TS model. We present comparative performance measures of GK algorithms with two clustering algorithms: (i)...
Combining LiDAR and camera data has shown potential in enhancing short-distance object detection autonomous driving systems. Yet, the fusion encounters difficulties with extended distance due to contrast between LiDAR's sparse dense resolution of cameras. Besides, discrepancies two representations further complicate methods. We introduce AYDIV, a novel framework integrating tri-phase alignment process specifically designed enhance long-distance even amidst discrepancies. AYDIV consists...