Abstract

Text as signs, labels, or instructions is a critical element of real-world scenes as they can convey important contextual information. 3D representations such as 3D Gaussian Splatting (3DGS) struggle to preserve fine-grained text details, while achieving high visual fidelity. Small errors in textual element reconstruction can lead to significant semantic loss. We propose STRinGS, a text-aware, selective refinement framework to address this issue for 3DGS reconstruction. Our method treats text and non-text regions separately, refining text regions first and merging them with non-text regions later for full-scene optimization. STRinGS produces sharp, readable text even in challenging configurations. We introduce a text readability measure OCR Character Error Rate (CER) to evaluate the efficacy on text regions. STRinGS results in a 63.6% relative improvement over 3DGS at just 7K iterations. We also introduce a curated dataset STRinGS-360 with diverse text scenarios to evaluate text readability in 3D reconstruction. Our method and dataset together push the boundaries of 3D scene understanding in text-rich environments, paving the way for more robust text-aware reconstruction methods.

Abstract

Scene understanding and reasoning has been a fundamental problem in 3D computer vision, requiring models to identify objects, their properties, and spatial or comparative relationships among the objects. Existing approaches enable this by creating scene graphs using multiple inputs such as 2D images, depth maps, object labels, and annotated relationships from specific reference view. However, these methods often struggle with generalization and produce inaccurate spatial relationships like "left/right", which become inconsistent across different viewpoints. To address these limitations, we propose Viewpoint-Invariant ZerO-shot scene graph generation for 3D scene Reasoning (VIZOR). VIZOR is a training-free, end-to-end framework that constructs dense, viewpoint-invariant 3D scene graphs directly from raw 3D scenes. The generated scene graph is unambiguous, as spatial relationships are defined relative to each object’s front-facing direction, making them consistent regardless of the reference view. Furthermore, it infers open-vocabulary relationships that describe spatial and proximity relationships among scene objects without requiring annotated training data. We conduct extensive quantitative and qualitative evaluations to assess the effectiveness of VIZOR on scene graph generation and downstream tasks, such as query-based object grounding. VIZOR outperforms state-of-the-art methods, showing clear improvements in scene graph generation and achieving 22% and 4.81% gains in zero-shot grounding accuracy on the Replica and Nr3D datasets, respectively.

Abstract

Generating realistic human motions that naturally respond to both spoken language and physical objects is crucial for interactive digital experiences. Current methods, however, address speech-driven gestures or object interactions independently, limiting real-world applicability due to a lack of integrated, comprehensive datasets. To overcome this, we introduce InteracTalker, a novel framework that seamlessly integrates prompt-based object-aware interactions with co-speech gesture generation. We achieve this by employing a multi-stage training process to learn a unified motion, speech, and prompt embedding space. To support this, we curate a rich human-object interaction dataset, formed by augmenting an existing text-to-motion dataset with detailed object interaction annotations. Our framework utilizes a Generalized Motion Adaptation Module that enables independent training, adapting to the corresponding motion condition, which is then dynamically combined during inference. To address the imbalance between heterogeneous conditioning signals, we propose an adaptive fusion strategy, which dynamically reweights the conditioning signals during diffusion sampling. InteracTalker successfully unifies these previously separate tasks, outperforming prior methods in both co-speech gesture generation and object-interaction synthesis, outperforming gesture-focused diffusion methods, yielding highly realistic, object-aware full-body motions with enhanced realism, flexibility, and control.(https://sreeharirajan.github.io/projects/InteracTalker/)

Abstract

Early-stage fruit yield prediction plays a key role in supporting timely agronomic decisions, enhancing market planning, and empowering farmers with data-driven insights. Over the years, most approaches to yield estimation have focused on fruit counting techniques, typically performed just before harvest. While these methods have proven useful, they often come into play late in the cultivation cycle, limiting their impact on early planning and resource optimization. In this work, we introduce a comprehensive baseline framework for predicting mango yield at an earlier stage - during flowering - using image-based learning. Our contributions are twofold. (i) Our approach combines a SegFormer-based segmentation model with a regression pipeline to estimate yield from images, while also exploring the role of contextual features such as weather and scale. (ii) This work introduces a novel benchmark and an enriched dataset, paving the way for scalable, automated tools that can assist farmers and stakeholders in making proactive decisions throughout the mango growing season. Our work demonstrates that for multi-modal yield prediction, integrating features that complement visual representations (like scale) can be more impactful than using features with a stronger standalone linear correlation (like weather). Our single-image model, based on the SegFormer-B1 encoder, achieved a mean absolute error (MAE) of 7.68, R² of 0.76, and mean squared error (MSE) of 115.48. These results highlight the promise of vision-based models for yield estimation from early-stage flowering cues. To the best of our knowledge, this is the first work to address the prediction of mango yield using images from the flowering stage and weather data.

Abstract

Temporal Graph Neural Networks (TGNNs) are increasingly used in high-stakes domains, such as financial forecasting, recommendation systems, and fraud detection. However, their susceptibility to poisoning attacks poses a critical security risk. We introduce LORETTA (Low Resource Twophase Temporal Attack), a novel adversarial framework on Continuous-Time Dynamic Graphs, which degrades TGNN performance by an average of 29.47% across 4 widely benchmark datasets and 4 State-of-the-Art (SotA) models. LORETTA operates through a two-stage approach: (1) sparsify the graph by removing high-impact edges using any of the 16 tested temporal importance metrics, (2) strategically replace removed edges with adversarial negatives via LORETTA’s novel degree-preserving negative sampling algorithm. Our plug-and-play design eliminates the need for expensive surrogate models while adhering to realistic unnoticeability constraints. LORETTA degrades performance by upto 42.0% on MOOC, 31.5% on Wikipedia, 28.8% on UCI, and 15.6%on Enron. LORETTA outperforms 11 attack baselines, remains undetectable to 4 leading anomaly detection systems, and is robust to 4 SotA adversarial defense training methods, establishing its effectiveness, unnoticeability, and robustness.

Abstract

This study evaluated the performance of 36 Coupled Model Inter-comparison Project Phase 6 (CMIP6) Global Climate Models (GCMs) in simulating extreme precipitation conditions across India using an integrated bias correction and multi-criteria decision-making framework. Daily precipitation data spanning 1975-2014 from CMIP6 GCMs were bias-corrected against Indian Meteorological Department (IMD) gridded observations at 0.25° spatial resolution using three distinct methods: (i) monthly quantile mapping applying empirical cumulative distribution function transformation on monthly aggregated data, (ii) daily quantile mapping using complete daily time series with linear interpolation for intermediate quantiles, and (iii) ISIMIP3b parametric trend-preserving quantile mapping employing gamma distribution fitting with mixed trend preservation strategy. Three extreme precipitation indices were calculated from bias-corrected data: total annual precipitation on wet days (PRCPTOTAL), 99th percentile of daily precipitation (R99p), frequency of days exceeding 99th percentile (R99pf). Performance evaluation employed Taylor Skill Score incorporating correlation and variance ratios, Pearson correlation coefficient (PCC), root-mean-square error (RMSE), and normalised standard deviation against IMD reference data. For the R99p index, PCC values ranged from -0.045 to 0.628, RMSE varied from 29.86 to 58.23 mm, and Taylor Skill Scores ranged from 0.72 to 0.17 across all GCMs. For R99p frequency, PCC ranged from 0.242 to 0.663, RMSE varied from 0.526 to 1.380, and Taylor Skill Scores ranged from 0.83 to 0.59. Multi-criteria decision-making integration used Analytic Hierarchy Process (AHP) with expert judgment-based pairwise comparison matrices for criteria weighting, followed by Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) and VlseKriterijumska Optimizacija Kompromisno Resenje (VIKOR) ranking methods with entropy-based objective weighting. The comprehensive ranking revealed EC-Earth3-Veg as the best performing GCM with a combined TOPSIS score of 0.890, followed by EC-Earth3-CC (0.880) and EC-Earth3 (0.875). MIROC6 achieved the highest Taylor Skill Score (0.717) for R99p intensity simulation, while GFDL-CM4 demonstrated consistent performance across both intensity and frequency metrics.

Abstract

This study presents a comprehensive multi-criteria ranking framework for evaluating 36 CMIP6 global climate models based on their simulation of temperature extreme indices over the Indian subcontinent. The research focuses on comparative assessment of Ta95p (95th percentile, hot extremes) and Ta5p (5th percentile, cold extremes) across critical monsoon (JJAS) and post-monsoon (OND) seasons. The methodology employs a systematic data processing pipeline beginning with IMD observational data (0.25°×0.25° resolution, 1975-2014) as reference, followed by CMIP6 model data preprocessing including spatial cropping to Indian domain (6.5°N-38.5°N, 66.5°E-100.0°E), bilinear interpolation to IMD resolution, and quantile mapping bias correction. Seasonal temperature extreme indices are calculated using robust percentile estimation methods for both seasons. The ranking framework integrates Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) and VIKOR (VIseKriterijumska Optimizacija I Kompromisno Resenje) multi-criteria decision-making techniques with AHP-derived weights (Pearson Correlation Coefficient, PCC: 0.652, Root Mean Square Error, RMSE: 0.239, Standard Deviation Normalization StdNorm: 0.109). TOPSIS calculates distances to ideal positive and negative solutions in multi-dimensional performance space, while VIKOR employs compromise programming emphasizing balanced performance across criteria. Final rankings aggregate both methods using mean rank averaging. Comparative analysis reveals distinct seasonal and index-specific performance patterns: TaiESM1 and CMCC-CM2-SR5 dominate hot extreme (Ta95p) simulation in both seasons, while EC-Earth3-CC and GFDL-ESM4 excel in monsoon cold extremes (Ta5p), and IPSL-CM5A2-INCA leads post-monsoon cold extreme representation. Performance metrics demonstrate substantial inter-model variability with PCC ranging 0.791-0.994, RMSE spanning 0.290-1.989°C, and StdNorm varying 0.835-1.213 across different indices and seasons.

Abstract

This study presents a comprehensive analysis of a statistical downscaling framework developed to predict daily rainfall over India. Utilizing outputs from the Coupled Model Intercomparison Project Phase 6 (CMIP6) and the National Centers for Environmental Prediction (NCEP) reanalysis data, the study employs a hybrid machine learning approach combining a classification model for broad meteorological zones with a cell-by-cell regression pipeline. The performance of six regression models- Lasso, Ridge, MLP Regressor, Random Forest, XGBoost, and Support Vector Regressor (SVR)—were evaluated using standard metrics. The findings reveal that while the framework successfully generates high-resolution rainfall projections, the models demonstrate limited explanatory power, with R² values ranging from 0.03 to 0.154. A critical discrepancy between the Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) indicates a significant failure to accurately predict extreme rainfall events, a key limitation for practical applications. The analysis concludes that while the framework provides a valuable baseline, a move towards more sophisticated architectures, such as deep learning and hybrid statistical-dynamical methods, is essential to capture the complex, non-linear spatiotemporal dynamics of precipitation.

Abstract

The relationship between tokenizer algorithm (e.g., Byte-Pair Encoding (BPE), Unigram), morphological alignment, tokenization quality (e.g., compression efficiency), and downstream performance remains largely unclear, particularly for languages with complex morphology. In this paper, we conduct a comprehensive evaluation of tokenizers using small-sized BERT models -- from pre-training through fine-tuning -- for Telugu (agglutinative), along with preliminary evaluation in Hindi (primarily fusional with some agglutination) and English (fusional). To evaluate morphological alignment of tokenizers in Telugu, we create a dataset containing gold morpheme segmentations of 600 derivational and 7000 inflectional word forms. Our experiments reveal two key findings for Telugu. First, the choice of tokenizer algorithm is the most significant factor influencing performance, with Unigram-based tokenizers consistently outperforming BPE across most settings. Second, while better morphological alignment shows a moderate, positive correlation with performance on text classification and structure prediction tasks, its impact is secondary to the tokenizer algorithm. Notably, hybrid approaches that use morphological information for pre-segmentation significantly boost the performance of BPE, though not Unigram. Our results further showcase the need for comprehensive intrinsic evaluation metrics for tokenizers that could explain downstream performance trends consistently.

Abstract

There are several use-cases where the hardness of a surface can be an important parameter - construction, infrastructure, sports fields, and so on. At present, there are no low-cost, thin-film, and wearable-based solutions for this problem. This study presents the development and evaluation of an insole-based sensing system for detecting and characterizing variations in surface conditions. The system integrates velostat, a piezoresistive material, into an insole to accurately capture plantar pressure variations as users traverse diverse terrains. The pressure data are subsequently processed to generate a detailed map differentiating various surface types. Experimental validation demonstrates that the system is capable of successfully mapping two distinct surfaces with an accuracy of 88.8 %. Multiple machine learning algorithms were analyzed to optimize the system performance and validate its efficacy. This approach has significant implications for the maintenance of a sports field, the optimization of athlete performance, and the prevention of injury.