arXiv subjects

Sumit Sharma

Publications and source records attributed to Sumit Sharma.

18 recordsLinked to original sources

arXiv2026

VAANI: Capturing the language landscape for an inclusive digital India

Voice based technologies have the potential to bridge digital accessibility gaps; however, existing datasets fail to capture the linguistic and regional diversity of Indic languages. We present Project VAANI, a large scale multimodal dataset designed to represent India's linguistic landscape across 165 districts. Speech data is collected using image based prompts to elicit spontaneous responses, while images are curated through a separate pipeline covering diverse themes across regions. The dataset undergoes a rigorous multi stage quality control process, combining automated and manual evaluation to ensure high audio quality and transcription accuracy. We release approximately 289K images, 31,255 hours of speech, and 2,043 hours of transcribed audio spanning 105 languages from 28 states and 3 union territories. Many of these languages are represented at this scale for the first time, making VAANI a foundational resource for inclusive speech technology. The dataset enables the development of robust, multilingual, and multimodal models, and supports research in speech recognition, language understanding, and cross-modal learning for underrepresented languages.

eess.AS↗

arXiv2026

Deliberative multi-agent large language models improve clinical reasoning in ophthalmology

Large language models (LLMs) show potential for ophthalmic clinical reasoning, yet individual models risk introducing harm. We evaluated whether multi-agent LLM deliberative councils improve diagnostic performance and mitigate harm compared to individual LLMs. In a comparative cross-sectional study, we assessed 12 individual LLMs and three multi-agent councils on 100 ophthalmology clinical vignettes. Each council comprised four models assembled by type: proprietary flagship, proprietary fast, and open-source. Models independently answered a vignette, anonymously ranked one another's responses, and a designated chair synthesized all responses and peer reviews into a final answer. Councils consistently outperformed pooled individual models across all three tiers. Accuracy improved for proprietary flagship (95.0% vs 90.8%; risk difference [RD]: 4.25 [95% CI: 0.45, 8.05]), proprietary fast (96.0% vs 86.5%; RD: 9.50 [5.31, 13.59]), and open-source councils (91.0% vs 83.2%; RD: 7.75 [4.17, 11.33]). Harm rates declined for proprietary flagship (10.0% vs 22.5%; RD: -12.50 [-16.86, -8.14]), proprietary fast (16.0% vs 31.8%; RD: -15.75 [-21.49, -10.01]), and open-source councils (22.0% vs 38.5%; RD: -16.50 [-22.27, -10.73]). Coverage analysis revealed net positive gains for accuracy ({\Delta}Coverage: 4.4-9.8 percentage points) and safety ({\Delta}Coverage: 13.6-20.6), indicating councils recovered correct diagnoses and averted harm. Councils elevated correct diagnoses to higher rank positions; and produced more complete differentials and management plans (all P<.05). Harmful council responses showed reduced combined commission-and-omission errors and tended to be less severe. Structured deliberation via multi-agent LLM councils may enhance the reliability of LLM-assisted ophthalmic clinical reasoning.

cs.CY↗

arXiv2026

Exploring parameter-efficient fine-tuning (PEFT) of billion-parameter vision models with QLoRA and DoRA: insights into generalization for limited-data image classification under a 98:1 test-to-train regime

Automated behavior classification is essential for precision livestock farming but faces challenges of high computational costs and limited labeled data. This study systematically compared three approaches: training from scratch (ResNet-18, ViT-Small), frozen feature extraction, and parameter-efficient fine-tuning (PEFT) of the DINOv3 foundation model (6.7 billion parameters). We evaluated QLoRA and DoRA across multiple configurations varying rank (8, 16, 64) and target modules (q_proj versus all-linear layers). With 2,160 verified training images, we assessed generalization of our model on 211,800 test samples, which is essentially a 98:1 test-to-train ratio. Results demonstrated that PEFT substantially outperformed alternatives, where the best QLoRA configuration (all-linear layers and rank=64) achieved 83.16% test accuracy with only 2.72% parameters (183.0M) in 5.8 hours, compared to 72.87% for ResNet-18 (16.8 hours), 61.91% for ViT-Small (18.7 hours), and 76.56% for frozen DINOv3 (17.5 hours). DoRA achieved comparable accuracy (83.14%) but with longer training time (11.0 hours). Notably, increasing adapter capacity consistently improved generalization while simultaneously not causing overfitting: reducing rank from 16 to 8 decreased test accuracy from 78.38% to 77.17%, while expanding from q_proj-only to all-linear layers with rank=64 improved accuracy from 78.38% to 83.16%. This suggests underfitting, instead of overfitting, is the primary challenge when adapting foundation models to agricultural imagery. Our findings provide guidelines for deploying billion-parameter vision models with PEFT in agricultural livestock applications.

cs.CV↗

arXiv2025

Hydration Free Energies of Linear Alkanes: Systematic Deviations in Common Water Models and Their Correction

Common force fields overestimate the hydration free energies of hydrophobic solutes, leading to an exaggerated hydrophobic effect. We compute the hydration free energies of linear alkanes from methane to eicosane (C${20}$H${42}$) using free energy perturbation with various three-site (SPC/E, OPC3) and four-site (TIP4P/2005, OPC) water models in combination with the TraPPE-UA alkane force field. All water models overestimate hydration free energies, although the four-site models perform better than the three-site ones. Using alkane cavity free energies, we reparameterize the alkane-water Lennard-Jones well depth to bring simulation results in agreement with experimental and group-contribution estimates at 300 K. The reparameterized models significantly improve agreement with experiments across temperatures (290--350 K). We also show that the General Amber Force Field (GAFF) with TIP4P/2005 water provides closer agreement with experimental hydration free energies than the original TraPPE-UA/TIP4P/2005 combination. Finally, we show that applying a shifted Lennard-Jones potential introduces systematic deviations in the hydration free energies.

cond-mat.soft↗

arXiv2025

Light Field Super-Resolution: A Critical Review on Challenges and Opportunities

Advances in portability and low cost of plenoptic cameras have revived interest in light field imaging. Light-field imaging has evolved into a technology that enables us to capture richer visual information. This high-dimensional representation of visual data provides a powerful way to understand the scene, with remarkable improvement in traditional computer vision problems such as depth sensing , post-capture refocusing , material classification, segmentation, and video stabilization. Capturing light fields with high spatial-angular resolution and capturing light field video at high frame rates remains a major challenge due to the limited resolution of the sensors, with limited processing speed. In this paper, we presented an extensive literature review of light field acquisition techniques, challenges associated with different capturing methodology and algorithms proposed for light-field super-resolution, in order to deal with spatial-angular resolution trade-off issue.

eess.IV↗

arXiv2025

A Computer Vision Pipeline for Individual-Level Behavior Analysis: Benchmarking on the Edinburgh Pig Dataset

Animal behavior analysis plays a crucial role in understanding animal welfare, health status, and productivity in agricultural settings. However, traditional manual observation methods are time-consuming, subjective, and limited in scalability. We present a modular pipeline that leverages open-sourced state-of-the-art computer vision techniques to automate animal behavior analysis in a group housing environment. Our approach combines state-of-the-art models for zero-shot object detection, motion-aware segmentation and tracking, and advanced feature extraction using vision transformers for robust behavior recognition. The pipeline addresses challenges including animal occlusions and group housing scenarios, as demonstrated in indoor pig monitoring. We validated our system on the Edinburgh Pig Behavior Video Dataset for multiple behavioral tasks. Our temporal model achieved 94.2% overall accuracy, representing a 21.2 percentage point improvement over existing methods. The pipeline demonstrated robust tracking capabilities with a 93.3% identity preservation (IDF1) score and an 89.3% average precision (AP) for object detection. The modular design suggests potential for adaptation to other contexts, though further validation across species would be required. The open-source implementation provides a scalable solution for behavior monitoring, contributing to precision pig farming and welfare assessment through automated, objective, and continuous analysis.

cs.CV↗

arXiv2025

Performance of GPT-5 Frontier Models in Ophthalmology Question Answering

Large language models (LLMs) such as GPT-5 integrate advanced reasoning capabilities that may improve performance on complex medical question-answering tasks. For this latest generation of reasoning models, the configurations that maximize both accuracy and cost-efficiency have yet to be established. We evaluated 12 configurations of OpenAI's GPT-5 series (three model tiers across four reasoning effort settings) alongside o1-high, o3-high, and GPT-4o, using 260 closed-access multiple-choice questions from the American Academy of Ophthalmology Basic Clinical Science Course (BCSC) dataset. The primary outcome was multiple-choice accuracy; secondary outcomes included head-to-head ranking via a Bradley-Terry model, rationale quality assessment using a reference-anchored, pairwise LLM-as-a-judge framework, and analysis of accuracy-cost trade-offs using token-based cost estimates. GPT-5-high achieved the highest accuracy (0.965; 95% CI, 0.942-0.985), outperforming all GPT-5-nano variants (P < .001), o1-high (P = .04), and GPT-4o (P < .001), but not o3-high (0.958; 95% CI, 0.931-0.981). GPT-5-high ranked first in both accuracy (1.66x stronger than o3-high) and rationale quality (1.11x stronger than o3-high). Cost-accuracy analysis identified several GPT-5 configurations on the Pareto frontier, with GPT-5-mini-low offering the most favorable low-cost, high-performance balance. These results benchmark GPT-5 on a high-quality ophthalmology dataset, demonstrate the influence of reasoning effort on accuracy, and introduce an autograder framework for scalable evaluation of LLM-generated answers against reference standards in ophthalmology.

cs.CL↗

arXiv2025

SPC to 3D: Novel View Synthesis from Binary SPC via I2I translation

Single Photon Avalanche Diodes (SPADs) represent a cutting-edge imaging technology, capable of detecting individual photons with remarkable timing precision. Building on this sensitivity, Single Photon Cameras (SPCs) enable image capture at exceptionally high speeds under both low and high illumination. Enabling 3D reconstruction and radiance field recovery from such SPC data holds significant promise. However, the binary nature of SPC images leads to severe information loss, particularly in texture and color, making traditional 3D synthesis techniques ineffective. To address this challenge, we propose a modular two-stage framework that converts binary SPC images into high-quality colorized novel views. The first stage performs image-to-image (I2I) translation using generative models such as Pix2PixHD, converting binary SPC inputs into plausible RGB representations. The second stage employs 3D scene reconstruction techniques like Neural Radiance Fields (NeRF) or Gaussian Splatting (3DGS) to generate novel views. We validate our two-stage pipeline (Pix2PixHD + Nerf/3DGS) through extensive qualitative and quantitative experiments, demonstrating significant improvements in perceptual quality and geometric consistency over the alternative baseline.

eess.IV↗

arXiv2024

Transforming Single Photon Camera Images to Color High Dynamic Range Images

Traditional CMOS sensors suffer from restricted dynamic range and sub optimal performance under extreme lighting conditions. They are affected by electronic noise in low light conditions and pixel saturation while capturing high illumination. Recent High Dynamic Range (HDR) Imaging methods, often designed for CMOS Sensors, attempt to address these issues by fusing multiple exposures. However, they frequently introduce artifacts like ghosting and light flickering in dynamic scenarios and non-uniform signal-to-noise ratio in extreme dynamic range conditions. Recently, Single Photon Avalanche Diodes (SPADs), also known as Single Photon Camera (SPC) sensors, have surpassed CMOS sensors due to their capability to capture individual photons with high timing precision. Unlike traditional digital cameras that first convert light energy into analog electrical currents and then digitize them, SPAD sensors perform direct photon detection, making them less susceptible to extreme illumination conditions. Their distinctive non-linear response curve aids in capturing photons across both low-light and high-illumination environments, making them particularly effective for High Dynamic Range Imaging. Despite their advantages, images from SPAD Sensors are often noisy and visually unappealing. To address these challenges, we evaluate state-of-the-art architectures for converting monochromatic SPAD images into Color HDR images at various resolutions. Our evaluation involves both qualitative and quantitative assessments of these architectures, focusing on their effectiveness in each stage of the conversion process.

eess.IV↗

arXiv2024

GN-FR:Generalizable Neural Radiance Fields for Flare Removal

Flare, an optical phenomenon resulting from unwanted scattering and reflections within a lens system, presents a significant challenge in imaging. The diverse patterns of flares, such as halos, streaks, color bleeding, and haze, complicate the flare removal process. Existing traditional and learning-based methods have exhibited limited efficacy due to their reliance on single-image approaches, where flare removal is highly ill-posed. We address this by framing flare removal as a multi-view image problem, taking advantage of the view-dependent nature of flare artifacts. This approach leverages information from neighboring views to recover details obscured by flare in individual images. Our proposed framework, GN-FR (Generalizable Neural Radiance Fields for Flare Removal), can render flare-free views from a sparse set of input images affected by lens flare and generalizes across different scenes in an unsupervised manner. GN-FR incorporates several modules within the Generalizable NeRF Transformer (GNT) framework: Flare-occupancy Mask Generation (FMG), View Sampler (VS), and Point Sampler (PS). To overcome the impracticality of capturing both flare-corrupted and flare-free data, we introduce a masking loss function that utilizes mask information in an unsupervised setting. Additionally, we present a 3D multi-view flare dataset, comprising 17 real flare scenes with 782 images, 80 real flare patterns, and their corresponding annotated flare-occupancy masks. To our knowledge, this is the first work to address flare removal within a Neural Radiance Fields (NeRF) framework.

cs.CV↗

arXiv2024

Force switching and potential shifting lead to significant cutoff dependence in alchemical free energies

The accurate treatment of long-range energy terms such as van der Waals interactions is crucial for reliable free energy calculations in molecular simulations. Methods like force switching, potential switching, potential shifting, and Ewald summation of van der Waals are commonly employed to smooth the truncation or otherwise manage these interactions at and beyond a cutoff distance, but their effects on free energy calculations are not always clear. In this study, we systematically explore the effects of these modifiers on the accuracy of free energy calculations using model systems: Lennard-Jones spheres, all-atom anthracene in water with GROMACS, and alkane chains in water with LAMMPS. Our results reveal that free energies of solvation using potential switching and particle-mesh Ewald summation of long-range Lennard-Jones are essentially independent of cutoff in solution, while force switching and potential shifting introduce cutoff-dependent behavior significant enough to affect the utility of the calculations.

physics.chem-ph↗

arXiv2024

MIS-ME: A Multi-modal Framework for Soil Moisture Estimation

Soil moisture estimation is an important task to enable precision agriculture in creating optimal plans for irrigation, fertilization, and harvest. It is common to utilize statistical and machine learning models to estimate soil moisture from traditional data sources such as weather forecasts, soil properties, and crop properties. However, there is a growing interest in utilizing aerial and geospatial imagery to estimate soil moisture. Although these images capture high-resolution crop details, they are expensive to curate and challenging to interpret. Imagine, an AI-enhanced software tool that predicts soil moisture using visual cues captured by smartphones and statistical data given by weather forecasts. This work is a first step towards that goal of developing a multi-modal approach for soil moisture estimation. In particular, we curate a dataset consisting of real-world images taken from ground stations and their corresponding weather data. We also propose MIS-ME - Meteorological & Image based Soil Moisture Estimator, a multi-modal framework for soil moisture estimation. Our extensive analysis shows that MIS-ME achieves a MAPE of 10.14%, outperforming traditional unimodal approaches with a reduction of 3.25% in MAPE for meteorological data and 2.15% in MAPE for image data, highlighting the effectiveness of tailored multi-modal approaches. Our code and dataset will be available at https://github.com/OSU-Complex-Systems/MIS-ME.git.

cs.CV↗

arXiv2023

Comprehensive Review on Semantic Information Retrieval and Ontology Engineering

Situation awareness is a crucial cognitive skill that enables individuals to perceive, comprehend, and project the current state of their environment accurately. It involves being conscious of relevant information, understanding its meaning, and using that understanding to make well-informed decisions. Awareness systems often need to integrate new knowledge and adapt to changing environments. Ontology reasoning facilitates knowledge integration and evolution, allowing for seamless updates and expansions of the ontology. With the consideration of above, we are providing a quick review on semantic information retrieval and ontology engineering to understand the emerging challenges and future research. In the review we have found that the ontology reasoning addresses the limitations of traditional systems by providing a formal, flexible, and scalable framework for knowledge representation, reasoning, and inference.

cs.IR↗

arXiv2023

Systematic design space exploration by learning the explored space using Machine Learning

Current practice in parameter space exploration in euclidean space is dominated by randomized sampling or design of experiment methods. The biggest issue with these methods is not keeping track of what part of parameter space has been explored and what has not. In this context, we utilize the geometric learning of explored data space using modern machine learning methods to keep track of already explored regions and samples from the regions that are unexplored. For this purpose, we use a modified version of a robust random-cut forest along with other heuristic-based approaches. We demonstrate our method and its progression in two-dimensional Euclidean space but it can be extended to any dimension since the underlying method is generic.

cs.LG↗

arXiv2021

Simulating Realistic MRI variations to Improve Deep Learning model and visual explanations using GradCAM

In the medical field, landmark detection in MRI plays an important role in reducing medical technician efforts in tasks like scan planning, image registration, etc. First, 88 landmarks spread across the brain anatomy in the three respective views -- sagittal, coronal, and axial are manually annotated, later guidelines from the expert clinical technicians are taken sub-anatomy-wise, for better localization of the existing landmarks, in order to identify and locate the important atlas landmarks even in oblique scans. To overcome limited data availability, we implement realistic data augmentation to generate synthetic 3D volumetric data. We use a modified HighRes3DNet model for solving brain MRI volumetric landmark detection problem. In order to visually explain our trained model on unseen data, and discern a stronger model from a weaker model, we implement Gradient-weighted Class Activation Mapping (Grad-CAM) which produces a coarse localization map highlighting the regions the model is focusing. Our experiments show that the proposed method shows favorable results, and the overall pipeline can be extended to a variable number of landmarks and other anatomies.

eess.IV↗

arXiv2015

Pathways to dewetting in hydrophobic confinement

Liquid water can become metastable with respect to its vapor in hydrophobic confinement. The resulting dewetting transitions are often impeded by large kinetic barriers. According to macroscopic theory, such barriers arise from the free energy required to nucleate a critical vapor tube that spans the region between two hydrophobic surfaces - tubes with smaller radii collapse, whereas larger ones grow to dry the entire confined region. Using extensive molecular simulations of water between two nanoscopic hydrophobic surfaces, in conjunction with advanced sampling techniques, here we show that for inter-surface separations that thermodynamically favor dewetting, the barrier to dewetting does not correspond to the formation of a (classical) critical vapor tube. Instead, it corresponds to an abrupt transition from an isolated cavity adjacent to one of the confining surfaces to a gap-spanning vapor tube that is already larger than the critical vapor tube anticipated by macroscopic theory. Correspondingly, the barrier to dewetting is also smaller than the classical expectation. We show that the peculiar nature of water density fluctuations adjacent to extended hydrophobic surfaces - namely, the enhanced likelihood of observing low-density fluctuations relative to Gaussian statistics - facilitates this non-classical behavior. By stabilizing isolated cavities relative to vapor tubes, enhanced water density fluctuations thus stabilize novel pathways, which circumvent the classical barriers and offer diminished resistance to dewetting. Our results thus suggest a key role for fluctuations in speeding up the kinetics of numerous phenomena ranging from Cassie-Wenzel transitions on superhydrophobic surfaces, to hydrophobically-driven biomolecular folding and assembly.

cond-mat.soft↗

arXiv2012

Age Based User Interface in Mobile Operating System

This paper proposes the creation of different interfaces in the mobile operating system for different age groups. The different age groups identified are kids, elderly people and all others. The motive behind creating different interfaces is to make the smartphones of today's world usable to all age groups.

cs.OS↗

arXiv2011

Evaporation Rate of Water in Hydrophobic Confinement

The drying of hydrophobic cavities is believed to play an important role in biophysical phenomena such as the folding of globular proteins, the opening and closing of ligand-gated ion channels, and ligand binding to hydrophobic pockets. We use forward flux sampling, a molecular simulation technique, to compute the rate of capillary evaporation of water confined between two hydrophobic surfaces separated by nanoscopic gaps, as a function of gap, surface size and temperature. Over the range of conditions investigated (gaps between 9 and 14 {\AA} and surface areas between 1 and 9 nm^2) the free energy barrier to evaporation scales linearly with the gap between hydrophobic surfaces, suggesting that line tension makes the predominant contribution to the free energy barrier. The exponential dependence of the evaporation rate on the gap between confining surfaces causes a ten order-of-magnitude decrease in the rate when the gap increases from 9 to 14 {\AA}. The computed free energy barriers are of the order of 50kT, and are predominantly enthalpic. Evaporation rates per unit area are found to be two orders of magnitude faster in confinement by the larger (9 nm^2) than by the smaller (1nm^2) surfaces considered here, at otherwise identical conditions. We show that this is a direct consequence of the dependence of hydrophobic hydration on the size of solvated objects. For sufficiently large surfaces, the critical nucleus for the evaporation process is a gap-spanning cylindrical vapor tube.

cond-mat.soft↗