arXiv subjects

Charalambos Poullis

Publications and source records attributed to Charalambos Poullis.

At least 19 recordsLinked to original sources

arXiv2026

OTT3R: Multi-View 3D Reconstruction and Fast Dataset Generation at 1% Compute

Feed-forward 3D reconstruction models have achieved impressive performance by scaling model and dataset size, but their cost excludes most research groups and precludes edge deployment. Additionally, generating 3D supervision without sensors still relies on slow, unreliable Structure-from-Motion, as the community lacks a COLMAP-like system for neural 3D pseudo-label generation. We present OTT3R (RGB-Only Tiny Transformer for 3D Reconstruction), a knowledge distillation framework that addresses both problems on a single workstation equipped with 2 GPUs. Distilling $π^3$ (959M parameters) into a 102M-parameter student yields 9.4$\times$ compression and up to 7$\times$ faster inference, trained at 1.6% of VGGT's training compute. An integrated pseudo-label pipeline offers a reliable, high-throughput alternative to COLMAP, generating dense per-pixel point maps and SE(3) camera poses for a 667K-image corpus in 3.5 hours on two commodity GPUs and succeeding on every sequence we tested, including those where COLMAP fails. The general student tracks the teacher on in-distribution monocular depth and, zero-shot, outperforms COLMAP on 7-Scenes and on DTU completion, but it does not replace the teacher on out-of-distribution multi-view geometry. The deployable artifact is the domain-specialized student: after specialization at 0.2% compute, it is 4$\times$ more accurate than COLMAP on 7-Scenes at 980$\times$ throughput, with near-teacher completion. Code is available at https://github.com/TheFourthKaramazov/OTT3R

cs.CV↗

arXiv2026

SAGE-OR: Semi-supervised Adaptive Scene Graph Generation for Operating Rooms

Current surgical scene graph generation methods depend on dense multi-modal supervision and specialized hardware (synchronized RGB-D sensors, calibration rigs), making dataset construction expensive and restricting all existing benchmarks to simulated environments. We propose SAGE-OR, a feature-centric framework that replaces the traditional detect-then-reason paradigm with a decoupled representation-reasoning paradigm in which localization is derived from frozen foundation models, encoded implicitly in pre-computed features, and used without any localization supervision, while a lightweight graph transformer performs relational reasoning over cached features. We employ a semi-supervised formulation with general-purpose segmentation prompts to eliminate localization supervision while enabling unsupervised context augmentation through additional prompt-driven entities, such as hands, which are absent from annotations. General-purpose prompts are used to induce near-perfect recall, while precision is delegated to downstream attention-based reasoning, enabling simple adaptation to new entities via prompt-level modification. This design enables a lightweight 15M-parameter graph transformer that trains in 1.4 hours and runs relational inference at $\sim$1ms per frame with peak memory under 2GB, suitable for edge hardware used in the operating room; feature extraction runs offline as a separate caching stage (4.27s per frame). On the 4D-OR benchmark, the core model achieves 76% F1, matching the fully supervised 4D-OR baseline while eliminating all localization annotations, and unsupervised hand augmentation raises this to 86%, within 4 points of state-of-the-art (SOTA) methods requiring dense multi-modal supervision, providing a practical pathway for adaptation to new surgical settings without annotation other than relationship and class labels.

cs.CV↗

arXiv2026

Data Leakage Detection and De-duplication in Large Scale Geospatial Image Datasets

In our study, we conducted a comprehensive analysis of three widely used datasets in the domain of building footprint extraction using deep neural networks: the INRIA Aerial Image Labelling dataset, SpaceNet 2: Building Detection v2, and the AICrowd Mapping Challenge datasets. Our experiments revealed several issues in the AICrowd Mapping Challenge dataset, where nearly 90% (about 250k) of the training split images had identical copies, indicating a high level of duplicate data. Additionally, we found that approximately 56k of the 60k images in the validation split were also present in the training split, amounting to a 93% data leakage. Furthermore, we present a data validation pipeline to address these issues of duplication and data leakage, which hinder the performance of models trained on such datasets. Employing perceptual hashing techniques, this pipeline is designed for efficient de-duplication and leakage identification. It aims to thoroughly evaluate the quality of datasets before their use, thereby ensuring the reliability and robustness of the trained models. Our code is available at https://github.com/yeshwanth95/Hash_and_search .

cs.CV↗

arXiv2026

Designing Fatigue-Aware VR Interfaces via Biomechanical Models

Prolonged mid-air interaction in virtual reality (VR) causes arm fatigue and discomfort, negatively affecting user experience. Incorporating ergonomic considerations into VR user interface (UI) design typically requires extensive human-in-the-loop evaluation. Although biomechanical models have been used to simulate human behavior in HCI tasks, their application as surrogate users for ergonomic VR UI design remains underexplored. We propose a hierarchical reinforcement learning framework that leverages biomechanical user models to evaluate and optimize VR interfaces for mid-air interaction. A motion agent is trained to perform button-press tasks in VR under sequential conditions, using realistic movement strategies and estimating muscle-level effort via a validated three-compartment control with recovery (3CC-r) fatigue model. The simulated fatigue output serves as feedback for a UI agent that optimizes UI element layout via reinforcement learning (RL) to minimize fatigue. We compare the RL-optimized layout against a manually-designed centered baseline and a Bayesian optimized baseline. Results show that fatigue trends from the biomechanical model align with human user data. Moreover, the RL-optimized layout using simulated fatigue feedback produced significantly lower perceived fatigue in a follow-up human study. We further demonstrate the framework's extensibility via a simulated case study on longer sequential tasks with non-uniform interaction frequencies. To our knowledge, this is the first work using simulated biomechanical muscle fatigue as a direct optimization signal for VR UI layout design. Our findings highlight the potential of biomechanical user models as effective surrogate tools for ergonomic VR interface design, enabling efficient early-stage iteration with less reliance on extensive human participation.

cs.HC↗

arXiv2026

Distill3R: A Pipeline for Democratizing 3D Foundation Models on Commodity Hardware

While multi-view 3D reconstruction has shifted toward large-scale foundation models capable of inferring globally consistent geometry, their reliance on massive computational clusters for training has created a significant barrier to entry for most academic laboratories. To bridge this compute divide, we introduce Distill3R, a framework designed to distill the geometric reasoning of 3D foundation models into compact students fully trainable on a single workstation. Our methodology centers on two primary innovations: (1) an offline caching pipeline that decouples heavy teacher inference from the training loop through compressed supervision signals, and (2) a confidence-aware distillation loss that leverages teacher uncertainty to enable training on commodity hardware. We propose a 72M-parameter student model which achieves a 9x reduction in parameters and a 5x inference speedup compared to its 650M-parameter teacher. The student is fully trainable in under 3 days on a single workstation, whereas its teacher requires massive GPU clusters for up to a week. We demonstrate that the student preserves the structural consistency and qualitative geometric understanding required for functional 3D awareness. By providing a reproducible, single-workstation training recipe, Distill3R serves as an exploratory entry point for democratized 3D vision research and efficient edge deployment. This work is not intended to compete with state-of-the-art foundation models, but to provide an accessible research baseline for laboratories without access to large-scale compute to train and specialize models on their own domain-specific data at minimal cost.

cs.CV↗

arXiv2025

Extreme Views: 3DGS Filter for Novel View Synthesis from Out-of-Distribution Camera Poses

When viewing a 3D Gaussian Splatting (3DGS) model from camera positions significantly outside the training data distribution, substantial visual noise commonly occurs. These artifacts result from the lack of training data in these extrapolated regions, leading to uncertain density, color, and geometry predictions from the model. To address this issue, we propose a novel real-time render-aware filtering method. Our approach leverages sensitivity scores derived from intermediate gradients, explicitly targeting instabilities caused by anisotropic orientations rather than isotropic variance. This filtering method directly addresses the core issue of generative uncertainty, allowing 3D reconstruction systems to maintain high visual fidelity even when users freely navigate outside the original training viewpoints. Experimental evaluation demonstrates that our method substantially improves visual quality, realism, and consistency compared to existing Neural Radiance Field (NeRF)-based approaches such as BayesRays. Critically, our filter seamlessly integrates into existing 3DGS rendering pipelines in real-time, unlike methods that require extensive post-hoc retraining or fine-tuning. Code and results at https://damian-bowness.github.io/EV3DGS

cs.CV↗

arXiv2025

Fast Self-Supervised depth and mask aware Association for Multi-Object Tracking

Multi-object tracking (MOT) methods often rely on Intersection-over-Union (IoU) for association. However, this becomes unreliable when objects are similar or occluded. Also, computing IoU for segmentation masks is computationally expensive. In this work, we use segmentation masks to capture object shapes, but we do not compute segmentation IoU. Instead, we fuse depth and mask features and pass them through a compact encoder trained self-supervised. This encoder produces stable object representations, which we use as an additional similarity cue alongside bounding box IoU and re-identification features for matching. We obtain depth maps from a zero-shot depth estimator and object masks from a promptable visual segmentation model to obtain fine-grained spatial cues. Our MOT method is the first to use the self-supervised encoder to refine segmentation masks without computing masks IoU. MOT can be divided into joint detection-ReID (JDR) and tracking-by-detection (TBD) models. The latter are computationally more efficient. Experiments of our TBD method on challenging benchmarks with non-linear motion, occlusion, and crowded scenes, such as SportsMOT and DanceTrack, show that our method outperforms the TBD state-of-the-art on most metrics, while achieving competitive performance on simpler benchmarks with linear motion, such as MOT17.

cs.CV↗

arXiv2025

Depth-Aware Scoring and Hierarchical Alignment for Multiple Object Tracking

Current motion-based multiple object tracking (MOT) approaches rely heavily on Intersection-over-Union (IoU) for object association. Without using 3D features, they are ineffective in scenarios with occlusions or visually similar objects. To address this, our paper presents a novel depth-aware framework for MOT. We estimate depth using a zero-shot approach and incorporate it as an independent feature in the association process. Additionally, we introduce a Hierarchical Alignment Score that refines IoU by integrating both coarse bounding box overlap and fine-grained (pixel-level) alignment to improve association accuracy without requiring additional learnable parameters. To our knowledge, this is the first MOT framework to incorporate 3D features (monocular depth) as an independent decision matrix in the association step. Our framework achieves state-of-the-art results on challenging benchmarks without any training nor fine-tuning. The code is available at https://github.com/Milad-Khanchi/DepthMOT

cs.CV↗

arXiv2025

DSV-LFS: Unifying LLM-Driven Semantic Cues with Visual Features for Robust Few-Shot Segmentation

Few-shot semantic segmentation (FSS) aims to enable models to segment novel/unseen object classes using only a limited number of labeled examples. However, current FSS methods frequently struggle with generalization due to incomplete and biased feature representations, especially when support images do not capture the full appearance variability of the target class. To improve the FSS pipeline, we propose a novel framework that utilizes large language models (LLMs) to adapt general class semantic information to the query image. Furthermore, the framework employs dense pixel-wise matching to identify similarities between query and support images, resulting in enhanced FSS performance. Inspired by reasoning-based segmentation frameworks, our method, named DSV-LFS, introduces an additional token into the LLM vocabulary, allowing a multimodal LLM to generate a "semantic prompt" from class descriptions. In parallel, a dense matching module identifies visual similarities between the query and support images, generating a "visual prompt". These prompts are then jointly employed to guide the prompt-based decoder for accurate segmentation of the query image. Comprehensive experiments on the benchmark datasets Pascal-$5^{i}$ and COCO-$20^{i}$ demonstrate that our framework achieves state-of-the-art performance-by a significant margin-demonstrating superior generalization to novel classes and robustness across diverse scenarios. The source code is available at \href{https://github.com/aminpdik/DSV-LFS}{https://github.com/aminpdik/DSV-LFS}

cs.CV↗

arXiv2024

Pix2Poly: A Sequence Prediction Method for End-to-end Polygonal Building Footprint Extraction from Remote Sensing Imagery

Extraction of building footprint polygons from remotely sensed data is essential for several urban understanding tasks such as reconstruction, navigation, and mapping. Despite significant progress in the area, extracting accurate polygonal building footprints remains an open problem. In this paper, we introduce Pix2Poly, an attention-based end-to-end trainable and differentiable deep neural network capable of directly generating explicit high-quality building footprints in a ring graph format. Pix2Poly employs a generative encoder-decoder transformer to produce a sequence of graph vertex tokens whose connectivity information is learned by an optimal matching network. Compared to previous graph learning methods, ours is a truly end-to-end trainable approach that extracts high-quality building footprints and road networks without requiring complicated, computationally intensive raster loss functions and intricate training pipelines. Upon evaluating Pix2Poly on several complex and challenging datasets, we report that Pix2Poly outperforms state-of-the-art methods in several vector shape quality metrics while being an entirely explicit method. Our code is available at https://github.com/yeshwanth95/Pix2Poly.

cs.CV↗

arXiv2024

Neural Real-Time Recalibration for Infrared Multi-Camera Systems

Currently, there are no learning-free or neural techniques for real-time recalibration of infrared multi-camera systems. In this paper, we address the challenge of real-time, highly-accurate calibration of multi-camera infrared systems, a critical task for time-sensitive applications. Unlike traditional calibration techniques that lack adaptability and struggle with on-the-fly recalibrations, we propose a neural network-based method capable of dynamic real-time calibration. The proposed method integrates a differentiable projection model that directly correlates 3D geometries with their 2D image projections and facilitates the direct optimization of both intrinsic and extrinsic camera parameters. Key to our approach is the dynamic camera pose synthesis with perturbations in camera parameters, emulating realistic operational challenges to enhance model robustness. We introduce two model variants: one designed for multi-camera systems with onboard processing of 2D points, utilizing the direct 2D projections of 3D fiducials, and another for image-based systems, employing color-coded projected points for implicitly establishing correspondence. Through rigorous experimentation, we demonstrate our method is more accurate than traditional calibration techniques with or without perturbations while also being real-time, marking a significant leap in the field of real-time multi-camera system calibration. The source code can be found at https://github.com/theICTlab/neural-recalibration

cs.CV↗

arXiv2024

HydroVision: LiDAR-Guided Hydrometric Prediction with Vision Transformers and Hybrid Graph Learning

Hydrometric forecasting is crucial for managing water resources, flood prediction, and environmental protection. Water stations are interconnected, and this connectivity influences the measurements at other stations. However, the dynamic and implicit nature of water flow paths makes it challenging to extract a priori knowledge of the connectivity structure. We hypothesize that terrain elevation significantly affects flow and connectivity. To incorporate this, we use LiDAR terrain elevation data encoded through a Vision Transformer (ViT). The ViT, which has demonstrated excellent performance in image classification by directly applying transformers to sequences of image patches, efficiently captures spatial features of terrain elevation. To account for both spatial and temporal features, we employ GRU blocks enhanced with graph convolution, a method widely used in the literature. We propose a hybrid graph learning structure that combines static and dynamic graph learning. A static graph, derived from transformer-encoded LiDAR data, captures terrain elevation relationships, while a dynamic graph adapts to temporal changes, improving the overall graph representation. We apply graph convolution in two layers through these static and dynamic graphs. Our method makes daily predictions up to 12 days ahead. Empirical results from multiple water stations in Quebec demonstrate that our method significantly reduces prediction error by an average of 10\% across all days, with greater improvements for longer forecasting horizons.

cs.CV↗

arXiv2023

TransGlow: Attention-augmented Transduction model based on Graph Neural Networks for Water Flow Forecasting

The hydrometric prediction of water quantity is useful for a variety of applications, including water management, flood forecasting, and flood control. However, the task is difficult due to the dynamic nature and limited data of water systems. Highly interconnected water systems can significantly affect hydrometric forecasting. Consequently, it is crucial to develop models that represent the relationships between other system components. In recent years, numerous hydrological applications have been studied, including streamflow prediction, flood forecasting, and water quality prediction. Existing methods are unable to model the influence of adjacent regions between pairs of variables. In this paper, we propose a spatiotemporal forecasting model that augments the hidden state in Graph Convolution Recurrent Neural Network (GCRN) encoder-decoder using an efficient version of the attention mechanism. The attention layer allows the decoder to access different parts of the input sequence selectively. Since water systems are interconnected and the connectivity information between the stations is implicit, the proposed model leverages a graph learning module to extract a sparse graph adjacency matrix adaptively based on the data. Spatiotemporal forecasting relies on historical data. In some regions, however, historical data may be limited or incomplete, making it difficult to accurately predict future water conditions. Further, we present a new benchmark dataset of water flow from a network of Canadian stations on rivers, streams, and lakes. Experimental results demonstrate that our proposed model TransGlow significantly outperforms baseline methods by a wide margin.

cs.LG↗

arXiv2023

ESPiM: Eye-Strain Probation Model, An Eye-Tracking Analysis Measure for Digital Displays

Eye-strain is a common issue among computer users due to the prolonged periods they spend working in front of digital displays. This can lead to vision problems, such as irritation and tiredness of the eyes and headaches. We propose the Eye-Strain Probation Model (ESPiM), a computational model based on eye-tracking data that measures eye-strain on digital displays based on the spatial properties of the user interface and display area for a required period of time. As well as measuring eye-strain, ESPiM can be applied to compare (a) different user interface designs, (b) different display devices, and (c) different interaction techniques. Two user studies were conducted to evaluate the effectiveness of ESPiM. The first was conducted in the form of an in-person study with an infrared eye-tracking sensor with 32 participants. The second was conducted in the form of an online study with a video-based eye-tracking technique via webcams on users' computers with 13 participants. Our analysis showed significantly different eye-strain patterns based on the video gameplay frequency of participants. Further, we found distinctive patterns among users on a regular 9-to-5 routine versus those with more flexible work hours in terms of (a) error rates and (b) reported eye-strain symptoms.

cs.HC↗

arXiv2022

End-to-End Multi-View Structure-from-Motion with Hypercorrelation Volumes

Image-based 3D reconstruction is one of the most important tasks in Computer Vision with many solutions proposed over the last few decades. The objective is to extract metric information i.e. the geometry of scene objects directly from images. These can then be used in a wide range of applications such as film, games, virtual reality, etc. Recently, deep learning techniques have been proposed to tackle this problem. They rely on training on vast amounts of data to learn to associate features between images through deep convolutional neural networks and have been shown to outperform traditional procedural techniques. In this paper, we improve on the state-of-the-art two-view structure-from-motion(SfM) approach of [11] by incorporating 4D correlation volume for more accurate feature matching and reconstruction. Furthermore, we extend it to the general multi-view case and evaluate it on the complex benchmark dataset DTU [4]. Quantitative evaluations and comparisons with state-of-the-art multi-view 3D reconstruction methods demonstrate its superiority in terms of the accuracy of reconstructions.

cs.CV↗

arXiv2022

Simpler is better: Multilevel Abstraction with Graph Convolutional Recurrent Neural Network Cells for Traffic Prediction

In recent years, graph neural networks (GNNs) combined with variants of recurrent neural networks (RNNs) have reached state-of-the-art performance in spatiotemporal forecasting tasks. This is particularly the case for traffic forecasting, where GNN models use the graph structure of road networks to account for spatial correlation between links and nodes. Recent solutions are either based on complex graph operations or avoiding predefined graphs. This paper proposes a new sequence-to-sequence architecture to extract the spatiotemporal correlation at multiple levels of abstraction using GNN-RNN cells with sparse architecture to decrease training time compared to more complex designs. Encoding the same input sequence through multiple encoders, with an incremental increase in encoder layers, enables the network to learn general and detailed information through multilevel abstraction. We further present a new benchmark dataset of street-level segment traffic data from Montreal, Canada. Unlike highways, urban road segments are cyclic and characterized by complicated spatial dependencies. Experimental results on the METR-LA benchmark highway and our MSLTD street-level segment datasets demonstrate that our model improves performance by more than 7% for one-hour prediction compared to the baseline methods while reducing computing resource requirements by more than half compared to other competing methods.

cs.LG↗

arXiv2022

Unsupervised Structure-Consistent Image-to-Image Translation

The Swapping Autoencoder achieved state-of-the-art performance in deep image manipulation and image-to-image translation. We improve this work by introducing a simple yet effective auxiliary module based on gradient reversal layers. The auxiliary module's loss forces the generator to learn to reconstruct an image with an all-zero texture code, encouraging better disentanglement between the structure and texture information. The proposed attribute-based transfer method enables refined control in style transfer while preserving structural information without using a semantic mask. To manipulate an image, we encode both the geometry of the objects and the general style of the input images into two latent codes with an additional constraint that enforces structure consistency. Moreover, due to the auxiliary loss, training time is significantly reduced. The superiority of the proposed model is demonstrated in complex domains such as satellite images where state-of-the-art are known to fail. Lastly, we show that our model improves the quality metrics for a wide range of datasets while achieving comparable results with multi-modal image generation techniques.

cs.CV↗

arXiv2022

Motion Estimation for Large Displacements and Deformations

Large displacement optical flow is an integral part of many computer vision tasks. Variational optical flow techniques based on a coarse-to-fine scheme interpolate sparse matches and locally optimize an energy model conditioned on colour, gradient and smoothness, making them sensitive to noise in the sparse matches, deformations, and arbitrarily large displacements. This paper addresses this problem and presents HybridFlow, a variational motion estimation framework for large displacements and deformations. A multi-scale hybrid matching approach is performed on the image pairs. Coarse-scale clusters formed by classifying pixels according to their feature descriptors are matched using the clusters' context descriptors. We apply a multi-scale graph matching on the finer-scale superpixels contained within each matched pair of coarse-scale clusters. Small clusters that cannot be further subdivided are matched using localized feature matching. Together, these initial matches form the flow, which is propagated by an edge-preserving interpolation and variational refinement. Our approach does not require training and is robust to substantial displacements and rigid and non-rigid transformations due to motion in the scene, making it ideal for large-scale imagery such as Wide-Area Motion Imagery (WAMI). More notably, HybridFlow works on directed graphs of arbitrary topology representing perceptual groups, which improves motion estimation in the presence of significant deformations. We demonstrate HybridFlow's superior performance to state-of-the-art variational techniques on two benchmark datasets and report comparable results with state-of-the-art deep-learning-based techniques.

cs.CV↗