The Amity Centre for Artificial Intelligence is a hub of innovation dedicated to advancing the frontiers of AI research across multiple disciplines. Our diverse team of researchers explores cutting-edge methodologies and applications in various domains including healthcare diagnostics, human-computer interaction, robotics, computer vision, and intelligent systems. Through our interdisciplinary approach, we address complex challenges by developing novel algorithms, frameworks, and practical solutions that have real-world impact.

Our research spans numerous AI subfields including Deep Learning, Transfer Learning, Brain-Computer Interfaces, Computer Vision, Internet of Things (IoT), Generative Adversarial Networks, and Human-Robot Interaction.

Below are some of our current research initiatives:

Depression Detection Using EEG Signals and Machine Learning Methods

  • Brief Description: Depression detection using EEG data involves analyzing electrical activity in the brain to identify patterns associated with depression. Machine learning and deep learning algorithms can then be used to accurately classify individuals as either depressed or not based on these patterns.
  • Lead: Dr. Karnati Mohan
  • Area: Brain Computer Interface & Machine Learning

Parkinson's Disease Detection Using EEG Signals and Deep Learning Methods

  • Brief Description: Parkinson's disease detection using EEG data and deep learning involves analyzing brain wave patterns to identify signs of the disease. Deep learning algorithms can be trained on EEG data to accurately classify individuals as either healthy or with Parkinson's disease based on these patterns.
  • Lead: Dr. Karnati Mohan, Dr. Geet Sahu
  • Area: Deep Learning & Brain Computer Interface

Visibility Restoration of Hazy Images Using Deep Learning Methods

  • Brief Description: Image dehazing involves removing unwanted haze or fog from images to improve their clarity and visual quality. This can be achieved using various traditional and deep learning techniques, which help restore the original details and colors of the image.
  • Lead: Dr. Geet Sahu
  • Area: Computer Vision & Deep Learning

Alzheimer's Disease Detection Using Brain Computer Interface and Deep Learning

  • Brief Description: Alzheimer's disease detection using MRI images involves analyzing patterns in brain scans to identify markers of Alzheimer's disease. A deep learning model can then be trained to accurately classify and detect Alzheimer's patients at various stages of the disease.
  • Lead: Dr. Geet Sahu
  • Area: Deep Learning & Brain Computer Interface

Schizophrenia Disease Detection Using EEG Signals and Deep Learning

  • Brief Description: Schizophrenia detection using EEG involves analyzing electrical activity in the brain to identify patterns associated with schizophrenia, such as abnormalities in gamma waves, connectivity patterns, and event-related potentials. Deep learning algorithms are developed and trained specifically for accurate schizophrenia detection.
  • Lead: Dr. Geet Sahu, Dr. Karnati Mohan
  • Area: Deep Learning & Brain Computer Interface

Schizophrenia Disease Detection Using EEG Signals and Deep Learning

  • Brief Description: Schizophrenia detection using EEG involves analyzing electrical activity in the brain to identify patterns associated with schizophrenia, such as abnormalities in gamma waves, connectivity patterns, and event-related potentials. Deep learning algorithms are developed and trained specifically for accurate schizophrenia detection.
  • Lead: Dr. Geet Sahu, Dr. Karnati Mohan
  • Area: Deep Learning & Brain Computer Interface

Multi-modal hand gesture recognition for robotic control application

  • Brief Description: Video and time-series data has been recorded for a large set of hand motion activities. Novel machine learning and deep learning models are being developed for integrated processing of this multi-modal data in order to improve classification performance.
  • Lead: Dr Rinki Gupta
  • Area: Human Computer Interaction & Time Series

Handling Data Scarcity in Deep Learning

  • Brief Description: Deep learning models require large amount of data for parameter optimization. Other techniques for training the deep learning model with limited labelled data are being implemented and evaluated for classification task. These include data augmentation using generative adversarial networks and use of Siamese network for time series data.
  • Lead: Dr Rinki Gupta
  • Area: Deep Learning

IoT-enabled Hand Gesture recognition and its application

  • Brief Description: Hand gestures is one of the prominent way to express and convey any information. This involves multiple applications such as prosthetic control, virtual gaming, rehabilitation, sign language and much more. In this we intend to develop such a IoT enabled wearable sensor based system for personal interaction in either of the application as given above.
  • Lead: Ms. Sneha Sharma
  • Area: Human Computer Interaction

Automated depression detection using deep learning with EEG signals

  • Brief Description: Depression detection using EEG data involves analyzing electrical activity in the brain to identify patterns associated with depression. Machine learning and Deep learning algorithms can then be used to accurately classify individuals as either depressed or not based on these patterns.
  • Lead: Ms. Sneha Sharma, Dr Rinki Gupta
  • Area: Brain Computer Interface & Deep Learning

Ensemble based transfer learning methods for wearable hand gesture recognition system

  • Brief Description: Development of a novel ensemble-based transfer learning algorithm for wearable hand gesture recognition system, which uses small amount of labelled data from a new user along with labelled data from other users to train an ensemble of learners for predicting unlabeled data from the new user.
  • Lead: Ms. Sneha Sharma, Dr Rinki Gupta
  • Area: Human Computer Interaction & Machine Learning

Deep transfer learning for recognition of continuous hand gestures

  • Brief Description: The proposed deep learning model consists of convolutional neural network (CNN), two bidirectional long short-term memory (Bi-LSTM) layers and connectionist temporal classification (CTC) to enable end-to-end recognition without requiring the knowledge of sign boundaries.
  • Lead: Ms. Sneha Sharma, Dr Rinki Gupta
  • Area: Deep Learning

Breast Cancer Classification Using Lightweight Transformer Models

  • Brief Description: Utilizing efficient transformer architectures to accurately identify and classify breast cancer. These models provide a computationally lightweight solution, enabling faster inference and efficient deployment while maintaining high classification performance for early detection and diagnosis of breast cancer.
  • Lead: Dr. Ritesh Maurya
  • Area: Computer Vision & Deep Learning

Tomato plant diseased image generation using Conditional GAN and classification

  • Brief Description: Combines generative adversarial networks (GANs) with classification models to create synthetic images of diseased tomato plants. Helps in training robust disease classification models by augmenting the dataset with synthetic samples.
  • Lead: Dr. Ritesh Maurya
  • Area: GAN & Computer Vision

Retinal Blood Vessels Segmentation and Multi-Class Classification of Fundus Images

  • Brief Description: A new AI framework for segmenting retinal blood vessels and classifying fundus images using modified U-Net and hybrid deep learning architecture. Combines features from segmented and raw images for better multi-class classification.
  • Lead: Prof. M.K. Dutta
  • Area: Computer Vision & Deep Learning

Customized Deep Neural Networks for Agricultural Image Classification

  • Brief Description: Multi-model deep learning framework for fine-grained classification in agricultural images with low visual differences. Extracts meaningful features capturing discriminative semantic info for multi-class tasks.
  • Lead: Prof. M.K. Dutta
  • Area: Computer Vision & Deep Learning

AI-based automatic classification of lumbar spine intervertebral disc-slices

  • Brief Description: An AI architecture is developed for the automatic classification of MRI images of mid-height intervertebral disc-slices of the lumbar spine. A diverse dataset is used for training, and various image pre-processing, model optimization and overfitting-handling strategies are incorporated for better decision-making. The multi-layer deep learning architecture extracts relevant features from multi-class MRI data, fine-tuning iterations for proper weight adjustments.
  • Lead: Prof. M.K. Dutta
  • Area: Computer Vision & Deep Learning

AI for Detecting and Predicting Various Grades of Breast Cancer

  • Brief Description: A novel machine-learning-based approach is proposed for identifying breast cancer and its multi-grade classification using biomarker data. The ensemble-based model can recognize control subjects and subjects susceptible to breast cancer, categorizing them into four levels of grades. The model is fine-tuned using minority oversampling techniques to avoid overfitting.
  • Lead: Prof. M.K. Dutta
  • Area: Machine Learning

Ensemble ML & Evolutionary Algorithms for Liver Cancer Diagnosis

  • Brief Description: This retrospective cohort study develops a machine-learning-based system for liver cancer prediction and precise diagnosis. The system uses ensemble learning to analyze data from diverse subjects and biomarkers, with evolutionary algorithms for hyper-parameter selection and oversampling for class balance.
  • Lead: Prof. M.K. Dutta
  • Area: Machine Learning

AI-powered Support System for Visually Impaired Individuals

  • Brief Description: This work presents an AI-based wearable assistive device that analyzes visual and sensory information about objects and obstacles. It generates auditory information and audio warnings, providing affordable, compact support for the visually impaired.
  • Lead: Prof. M.K. Dutta
  • Area: Computer Vision & Deep Learning

Improved Ensemble-based Transfer Learning for ISL Recognition

  • Brief Description: This work proposes an ensemble of transfer learning methods using VGG16, ResNet50, and InceptionV3 models. Fine-tuned on Indian Sign Language datasets, the models' predictions are combined to improve recognition accuracy.
  • Lead: Dr. Sneha Sharma
  • Area: Human Computer Interaction & Deep Learning

Autistic Behavior Analysis Using Machine Learning

  • Brief Description: Fish movements are tracked under different conditions (control and intoxicated). Based on these parameters, a machine learning algorithm is trained to detect autistic-like behavior in the specimens.
  • Lead: Prof. M.K. Dutta
  • Area: Neural Science & Machine Learning

Training Pace Optimization for Low Frequency Time Series

  • Brief Description: This work introduces a training pace optimization algorithm and custom CNN architecture for analyzing low-frequency time series data. Adaptive learning and bias-correction methods improve training convergence.
  • Lead: Prof. M.K. Dutta
  • Area: Time Series & Deep Learning

Deep Neural Network for Detection of Pulmonary Sounds

  • Brief Description: Proposes an AI-based diagnostic framework for identifying pulmonary sounds using deep learning. The model is optimized using various training strategies to ensure adaptability and high performance.
  • Lead: Prof. M.K. Dutta
  • Area: Time Series & Deep Learning

Transformer-based Network for Microorganism Classification

  • Brief Description: Due to challenges in feature extraction from limited-size datasets, a hybrid architecture combining CNN and Vision Transformer is proposed for classifying environmental microorganisms via microscopic images.
  • Lead: Prof. M.K. Dutta
  • Area: Vision Transformer

AI based Predictive Model for Drug-Protein Binding Affinity Prediction using 1D-CNN

  • Brief Description: Predicting drug target binding affinity has huge relevance in modern drug discovery and repositioning. In silico models using deep learning can assist in identifying promising drug-target pairs. This study proposes a deep learning framework to support drug-target interaction prediction.
  • Lead: Prof. M. K. Dutta
  • Area: Time Series & Deep Learning

Skin Cancer Classification using Multi-model Deep Learning and Grey Wolf Optimization

  • Brief Description: This study proposes a novel framework combining multiple deep convolutional neural architectures and Grey Wolf Optimization to classify skin lesions. Information gain is used for feature selection, enhancing classification performance.
  • Lead: Dr. Ritesh Maurya
  • Area: Computer Vision & Deep Learning

Machine Learning in Risk Prediction of Type 2 Diabetes Mellitus

  • Brief Description: The study utilizes anthropometric, biochemical, and genetic parameters to identify T2DM risk factors. Machine learning and recursive feature elimination are used to build efficient predictive models.
  • Lead: Prof. M. K. Dutta
  • Area: Machine Learning

Optimization Framework for Enhanced Learning Stability on Large-Scale Datasets

  • Brief Description: This work proposes advanced optimization techniques for stable and fast learning with minimal hyper-parameter tuning. The framework emphasizes fewer parameters for better speed and reduced complexity.
  • Lead: Prof. M. K. Dutta
  • Area: Deep Learning

Deep Neural Network-Based Framework for WBC Analysis from Microscopic Images

  • Brief Description: A generic deep learning-based model is developed to analyze and classify microscopic blood cell images, addressing challenges like resolution differences, staining, and illumination variations.
  • Lead: Prof. M. K. Dutta
  • Area: Computer Vision & Deep Learning

Deep Learning-based Diagnosis of Brain Tumor

  • Brief Description: Development of deep learning-based AI architectures for diagnosing various categories of brain tumors.
  • Lead: Dr. Ritesh Maurya
  • Area: Computer Vision & Deep Learning

Deep CNN for Epilepsy Detection Using EEG Signals

  • Brief Description: Design of deep learning-based AI architectures to detect epilepsy through EEG signal analysis.
  • Lead: Dr. Karnati Mohan & Dr. Geet Sahu
  • Area: Time Series & Deep Learning

Facial Expression Recognition using Fractal Dimension and Deep CNNs

  • Brief Description: AI architectures based on deep learning and fractal features are designed for classifying facial expressions.
  • Lead: Dr. Karnati Mohan
  • Area: Computer Vision & Deep Learning

Wearable Sensor-Based Elderly Fall Detection and Monitoring System

  • Brief Description: A novel framework is developed to detect and monitor fall activities in elderly individuals using wearable sensor data and deep learning.
  • Lead: Dr. Sneha Sharma
  • Area: Human Computer Interaction & Deep Learning

Joint Angle Prediction in Robotics using Machine Learning methods

  • Brief Description: Machine Learning-Driven Inverse Kinematics and Joint Angle Prediction for Six-DoF Anthropomorphic Robots with Explainable AI
  • Lead: Dr. J.K. Rai & Dr. M.K. Dutta
  • Area: Machine Learning & Robotics

WBC classification using Deep Neural Network from microscopic images

  • Brief Description: Design of deep learning-based generic framework for segmentation and classification of microscopic white blood cell images.
  • Lead: Prof. M. K. Dutta
  • Area: Computer Vision & Deep Learning

Training pace optimization under noisy conditions for Time Series signals

  • Brief Description: Design of Robust Gradient Estimation using a Optimizer for Deep Neural Networks to Efficiently Handle Noisy Time Series Signals.
  • Lead: Prof. M. K. Dutta
  • Area: Deep Learning

Early Detection of Schizophrenia using Deep Neural Network

  • Brief Description: Design and Development of Deep Learning based Architectures for Classification of EEG Signals for Detection of Schizophrenia in Real Time.
  • Lead: Dr. Geet Sahu
  • Area: Time Series & Deep Learning

A Multimodality based depression Detection System

  • Brief Description: To design a robust neural network using the multimodal data to detect the stage of depression
  • Lead: Dr. Sneha Sharma
  • Area: Deep Learning

Deep Learning Model for Early Diagnosis of Alzheimer’s Disease

  • Brief Description: Design of Deep Learning Framework for Automatic Detection of Alzheimer’s Disease from EEG Signal.
  • Lead: Dr. Geet Sahu
  • Area: Deep Learning

Enhancing Safety and Durability: Deep Learning-Powered Detection of Cracks in Decks, Walls, and Pavements

  • Brief Description: This research leverages deep learning to detect cracks in essential infrastructure components. Using advanced neural networks and computer vision, it enables early detection, real-time monitoring, and cost-effective maintenance for safer urban environments.
  • Lead: Ayush Singh Rajput, Prof. M.K. Dutta, Dr. Priyanka Singh
  • Area: Deep Learning

Comprehensive Comparative Analysis of Machine Learning Algorithms for Predicting the Bond Strength of Concrete

  • Brief Description: This study explores various ML algorithms to predict concrete bond strength, critical for construction safety. A comparative analysis ensures the selection of optimal predictive methods for structural integrity.
  • Lead: Mr. Mayank Chaudhary, Dr. Priyanka Singh, Dr. Malay Kishor Dutta
  • Area: Machine Learning

Deep Learning Framework to Predict the Trend in Stock Market

  • Brief Description: Development of deep learning methods that predict financial market trends by automating statistical data and forecasting stock price movements.
  • Lead: Sneha Sharma
  • Area: Time Series Forecasting

Deep Learning-based Diagnosis of Brain Stroke

  • Brief Description: To develop deep learning based method for automatic detection of brain stroke using CT scans, enabling faster diagnosis.
  • Lead: Dr. Ritesh Maurya
  • Area: Deep Learning, Computer Vision

Pneumonia Detection using Deep Learning with Novel Loss Function

  • Brief Description: To develop a deep learning-based methodology for fast pneumonia diagnosis using X-ray images and a customized loss function.
  • Lead: Dr. Ritesh Maurya
  • Area: Deep Learning, Computer Vision