A5077592877- Domain Adaptation and Few-Shot Learning
- Radiomics and Machine Learning in Medical Imaging
- Remote-Sensing Image Classification
- Advanced Image and Video Retrieval Techniques
- Machine Learning and Data Classification
- Video Surveillance and Tracking Methods
- COVID-19 diagnosis using AI
- AI in cancer detection
- Machine Learning in Healthcare
- Image Retrieval and Classification Techniques
- Vehicle License Plate Recognition
- Medical Imaging and Analysis
- Respiratory viral infections research
- Multimodal Machine Learning Applications
- Gait Recognition and Analysis
- Computational Drug Discovery Methods
- Hate Speech and Cyberbullying Detection
- Biomedical Text Mining and Ontologies
- Advanced Neural Network Applications
- Medical Image Segmentation Techniques
- Rough Sets and Fuzzy Logic
- Human Pose and Action Recognition
- Anomaly Detection Techniques and Applications
- Colorectal Cancer Screening and Detection
- Appendicitis Diagnosis and Management
Jadavpur University
2022-2024
Indian Institute of Technology Bombay
2023-2024
Single-source open-domain generalization (SS-ODG) addresses the challenge of labeled source domains with supervision during training and unlabeled novel target testing. The domain includes both known classes from samples previously unseen classes. Existing techniques for SS-ODG primarily focus on calibrating source-domain classifiers to identify open in domain. However, these methods struggle visually fine-grained open-closed data, often misclassifying as closed-set Moreover, relying solely...
We focus on domain and class generalization problems in analyzing optical remote sensing images, using the large-scale pre-trained vision-language model (VLM), CLIP. While contrastively trained VLMs show impressive zero-shot performance, their effectiveness is limited when dealing with diverse domains during training testing. Existing prompt learning techniques overlook importance of incorporating content information into prompts, which results a drop performance while such multi-domain...
We introduce a new way to train Multi-Layer Perceptron (MLP) classify incomplete data. To achieve this, we an MLP using two-phased approach. In the first phase, complete create augmented dataset before second phase of training. For use non-missing data, delete each feature once, and then fill it some predefined points. After that, in retrain network dataset. The aim this type training is predict class label At time testing, when vector with missing value appears, initially impute points find...
We present a novel method for improving the adaptability of self-supervised (SSL) pre-trained models across different domains. Our approach uses synthetic images that are generated using an auxiliary diffusion model, namely InstructPix2Pix. More specifically, starting from real image, we prompt model to generate versions image in style target This allows us diverse set multi-domain share same semantics as images. Integrating these into training dataset enhances model's capacity generalize...
Proliferation of online political hate speech through social media has been a persisting problem and is being recently compounded by the arrival AI-boosted content. This can lead to wanton dissemination misinformation/disinformation cause extremist radicalisation or influence national electoral processes. Given high stakes negative impact, it becoming increasingly important address sensitive topic content moderation on platforms, debate dichotomy free versus harm. From that perspective,...
Unets have become the standard method for semantic segmentation of medical images, along with fully convolutional networks (FCN). Unet++ was introduced as a variant Unet, in order to solve some problems facing Unet and FCNs. provided an ensemble variable depth Unets, hence eliminating need professionals estimating best suitable task. While all its variants, including aimed at providing that were able train well without requiring large quantities annotated data, none them attempted eliminate...
The process of drug discovery has received a big boost from the use high-end computers and machine learning algorithms. There are about 38 new drugs introduced into market each year, making it difficult for physicians to keep up with latest advancements. In this work we introduce an end-to-end system that is able go through literature newly discovered store features in database. Then, accept symptoms patient physician recommend appropriate drugs, which may be prescribed patient, subject...
Abstract We introduced a new way to train classifier classify different data specifically incomplete data. Here, we using two-phase approach. In the first phase, complete Then, create set (augmented dataset) before second phase of training. After that, in retrain newly created set. At time testing, if feature vector with missing value appears, initially impute it various strategies and it. Then try find class label each trained classifier. If clas-sifier is augmented dataset, performance...
Single-source open-domain generalization (SS-ODG) addresses the challenge of labeled source domains with supervision during training and unlabeled novel target testing. The domain includes both known classes from samples previously unseen classes. Existing techniques for SS-ODG primarily focus on calibrating source-domain classifiers to identify open in domain. However, these methods struggle visually fine-grained open-closed data, often misclassifying as closed-set Moreover, relying solely...
We focus on domain and class generalization problems in analyzing optical remote sensing images, using the large-scale pre-trained vision-language model (VLM), CLIP. While contrastively trained VLMs show impressive zero-shot performance, their effectiveness is limited when dealing with diverse domains during training testing. Existing prompt learning techniques overlook importance of incorporating content information into prompts, which results a drop performance while such multi-domain...
Abdominal pain is one of the most common symptoms for a wide range conditions in children, under age 16 years. Due to limited ability X-ray distinguish between structures soft tissue, physicians often rely on Computed Tomography (CT) scan diagnose underlying cause abdominal pain. A CT exposes patient 70-150 times radiation used an X-ray. Moreover, scanning equipment not available low-resource countries, leading improper diagnosis and treatment. Children are more susceptible harmful effects...
The Syrian civil war has forced nearly 5.6 million people to seek refuge in nearby nations. Out of these, 2.7 are children and around 6 under the age five years old. Such experiences increase risk psychological trauma young and, if left untreated, might have an impact on rest their lives. Children like paint. Hence, paintings effective non invasive source information issues being faced by children. In this work, we propose a novel architecture that is able reliably detect with trauma. For...