A5076951754- Robotic Path Planning Algorithms
- Topic Modeling
- Multimodal Machine Learning Applications
- Robotics and Sensor-Based Localization
- Sentiment Analysis and Opinion Mining
- Evacuation and Crowd Dynamics
- Gaze Tracking and Assistive Technology
- Social Robot Interaction and HRI
- Stellar, planetary, and galactic studies
- Autonomous Vehicle Technology and Safety
- Robotics and Automated Systems
- Astro and Planetary Science
- Text and Document Classification Technologies
- Anomaly Detection Techniques and Applications
- Robotic Locomotion and Control
- E-commerce and Technology Innovations
- Software Testing and Debugging Techniques
- Adversarial Robustness in Machine Learning
- Advanced Steganography and Watermarking Techniques
- Human Pose and Action Recognition
- Transgenic Plants and Applications
- Computational Drug Discovery Methods
- Domain Adaptation and Few-Shot Learning
- Complex Systems and Time Series Analysis
- Music and Audio Processing
National Law University, Delhi
2025
Birla Institute of Technology and Science, Pilani
2022-2024
National Student Clearinghouse Research Center
2024
Massachusetts Institute of Technology
2023-2024
Indraprastha Institute of Information Technology Delhi
2023-2024
IIT@MIT
2024
LNM Institute of Information Technology
2023
Chandigarh University
2023
Guru Gobind Singh Indraprastha University
2023
University of Pennsylvania
2020
This paper explores the task Natural Language Understanding (NLU) by looking at duplicate question detection in Quora dataset. We conducted extensive exploration of dataset and used various machine learning models, including linear tree-based models. Our final finding was that a simple Continuous Bag Words neural network model had best performance, outdoing more complicated recurrent attention based also error analysis found some subjectivity labeling
A key challenge in off-road navigation is that even visually similar terrains or ones from the same semantic class may have substantially different traction properties. Existing work typically assumes no wheel slip uses expected for motion planning, where predicted trajectories provide a poor indication of actual performance if terrain has high uncertainty. In contrast, this proposes to analyze traversability with empirical distribution parameters unicycle dynamics, which can be learned by...
Heat stress (HS) is a major threat to crop productivity and expected be more frequent severe due climate change challenges. The predicted increase in global temperature requires us understand the dimensions of HS experienced by plants, particularly during reproductive stages, as majorly dependent on success plant reproduction. impact relatively less-studied than other abiotic stresses, such drought salinity. Plants have evolved diverse mechanisms perceive, transduce, respond, adapt at...
Traversing terrain with good traction is crucial for achieving fast off-road navigation. Instead of manually designing costs based on features, existing methods learn properties directly from data via self-supervision to automatically penalize trajectories moving through undesirable terrain, but challenges remain in properly quantifying and mitigating the risk due uncertainty learned models. To this end, we present evidential autonomy (EVORA), a unified framework uncertainty-aware model plan...
We present a catalog of 316 trans-Neptunian bodies detected by the Dark Energy Survey (DES). These objects include 245 discoveries DES (139 not previously published) in $\approx 60,000$ exposures from first four seasons survey ("Y4" data). The covers contiguous 5000 deg$^2$ southern sky $grizY$ optical/NIR filter set, with typical TNO this part being targeted $25-30$ Y4 exposures. describe processes for detection transient sources and linkage into orbits, which are made challenging absence...
A key challenge in fast ground robot navigation 3D terrain is balancing speed and safety. Recent work has shown that 2.5D maps (2D representations with additional information) are ideal for real-time safe planning. However, the prevalent approach of generating 2D occupancy grids through raytracing makes generated map unsafe to plan in, due inaccurate representation unknown space. Additionally, existing planners such as MPPI do not consider speeds known free space separately, leading slower...
Early crop yield prediction is crucial in agriculture for making administrative plans to ensure food security, post harvest management and distribution of a crop. Remote sensing data captured using various satellites provide reliable phenological information through surface reflectance bands. Other important factors, affecting include meteorological soil. The which we have used multimodal. It consists spatiotemporal (numeric) bands (satellite image), temporally static soil image) data. We...
Spatial cognition, as gained through the sense of vision, is one most important capabilities human beings. However, for visually impaired (VI), lack this perceptual capability poses great challenges in their life. Therefore, we have designed Point-to-Tell-and-Touch, a wearable system with an ergonomic human-machine interface, assisting VI active environmental exploration, particular focus on spatial intelligence and navigation to objects interest alien environment. Our key idea link visual...
Abstract: Music genre classification is a fundamental task in the field of music information retrieval (MIR) and has gained significant attention recent years due to rapid growth digital collections. This research paper presents comprehensive review application machine learning techniques for classification. We explore various methodologies, feature extraction techniques, algorithms used domain, highlighting their strengths, limitations, advancements. The objective this provide researchers...
Vision-Language Models (VLMs) such as CLIP are trained on large amounts of image-text pairs, resulting in remarkable generalization across several data distributions. However, cases, their expensive training and collection/curation costs do not justify the end application. This motivates a vendor-client paradigm, where vendor trains large-scale VLM grants only input-output access to clients pay-per-query basis black-box setting. The client aims minimize inference cost by distilling student...
This paper examines how AI has revolutionised drug development and medical research using the ChEMBL dataset. The primary study areas are AI-driven therapeutic target identification., computational approaches in development., repurposing for COVID-19 therapies., methods natural leather flaw detection. Target selection must balance novelty confidence., identification is considered. Structure-based virtual screening profound learning predictions of ligand properties activities considered...
In this project we analysed how much semantic information images carry, and value image data can add to sentiment analysis of the text associated with images. To better understand contribution from images, compared models which only made use data, combined both types. We also if approach could help classifiers generalize unknown sentiments.
To operate safely and efficiently, autonomous warehouse/delivery robots must be able to accomplish tasks while navigating in dynamic environments handling the large uncertainties associated with motions/behaviors of other and/or humans. A key scenario such is hallway problem, where same narrow corridor as human traffic going one or both directions. Traditionally, robot planners have tended focus on socially acceptable behavior at expense performance. This paper proposes a planner that aims...
31May 2017 SYMBOLIC CLASSIFICATION FOR MULTIVARIATE TIME SERIES. Amanpreet Singh , Dashmeet Kaur Sethi Karneet Lakshay Sharma and Poonam Narang. Btech (CSE), GTBIT, New Delhi. Ast. Professor(CSE/IT),
Social media users are increasingly using images and videos to communicate their feelings. However, researchers usually rely on textual data for sentiment analysis. In this paper, we focus Twitter posts having both text image analyze the match between sentiments, when predicted independently. We also propose a method predict of social post as collective entity from individual sentiments image. Comparing our proposed model naive approach where is assigned also, see an improvement around 20%...
OVSF codes in the WCDMA system provide facility for variable data rate support. The applications with different rates must be served fair manner. paper propose a single code and multi design so that tree is not over-served type of application. considers partitioning adaptive to user rates. based upon selection out number available options such usage leads most assignment. average blocking may significantly reduced, but individual high one or more schemes. Simulation results are presented...