arXiv subjects

Dimitrios Zarpalas

Publications and source records attributed to Dimitrios Zarpalas.

At least 19 recordsLinked to original sources

arXiv2026

Geo-NVS-w: Geometry-Aware Novel View Synthesis In-the-Wild with an SDF Renderer

We introduce Geo-NVS-w, a geometry-aware framework for high-fidelity novel view synthesis from unstructured, in-the-wild image collections. While existing in-the-wild methods already excel at novel view synthesis, they often lack geometric grounding on complex surfaces, sometimes producing results that contain inconsistencies. Geo-NVS-w addresses this limitation by leveraging an underlying geometric representation based on a Signed Distance Function (SDF) to guide the rendering process. This is complemented by a novel Geometry-Preservation Loss which ensures that fine structural details are preserved. Our framework achieves competitive rendering performance, while demonstrating a 4-5x reduction reduction in energy consumption compared to similar methods. We demonstrate that Geo-NVS-w is a robust method for in-the-wild NVS, yielding photorealistic results with sharp, geometrically coherent details.

cs.CV↗

arXiv2025

Image Valuation in NeRF-based 3D reconstruction

Data valuation and monetization are becoming increasingly important across domains such as eXtended Reality (XR) and digital media. In the context of 3D scene reconstruction from a set of images -- whether casually or professionally captured -- not all inputs contribute equally to the final output. Neural Radiance Fields (NeRFs) enable photorealistic 3D reconstruction of scenes by optimizing a volumetric radiance field given a set of images. However, in-the-wild scenes often include image captures of varying quality, occlusions, and transient objects, resulting in uneven utility across inputs. In this paper we propose a method to quantify the individual contribution of each image to NeRF-based reconstructions of in-the-wild image sets. Contribution is assessed through reconstruction quality metrics based on PSNR and MSE. We validate our approach by removing low-contributing images during training and measuring the resulting impact on reconstruction fidelity.

cs.CV↗

arXiv2025

Learning Encoding-Decoding Direction Pairs to Unveil Concepts of Influence in Deep Vision Networks

Empirical evidence shows that deep vision networks often represent concepts as directions in latent space with concept information written along directional components in the vector representation of the input. However, the mechanism to encode (write) and decode (read) concept information to and from vector representations is not directly accessible as it constitutes a latent mechanism that naturally emerges from the training process of the network. Recovering this mechanism unlocks significant potential to open the black-box nature of deep networks, enabling understanding, debugging, and improving deep learning models. In this work, we propose an unsupervised method to recover this mechanism. For each concept, we explain that under the hypothesis of linear concept representations, this mechanism can be implemented with the help of two directions: the first facilitating encoding of concept information and the second facilitating decoding. Unlike prior matrix decomposition, autoencoder, or dictionary learning methods that rely on feature reconstruction, we propose a new perspective: decoding directions are identified via directional clustering of activations, and encoding directions are estimated with signal vectors under a probabilistic view. We further leverage network weights through a novel technique, Uncertainty Region Alignment, which reveals interpretable directions affecting predictions. Our analysis shows that (a) on synthetic data, our method recovers ground-truth direction pairs; (b) on real data, decoding directions map to monosemantic, interpretable concepts and outperform unsupervised baselines; and (c) signal vectors faithfully estimate encoding directions, validated via activation maximization. Finally, we demonstrate applications in understanding global model behavior, explaining individual predictions, and intervening to produce counterfactuals or correct errors.

cs.CV↗

arXiv2025

Which Direction to Choose? An Analysis on the Representation Power of Self-Supervised ViTs in Downstream Tasks

Self-Supervised Learning (SSL) for Vision Transformers (ViTs) has recently demonstrated considerable potential as a pre-training strategy for a variety of computer vision tasks, including image classification and segmentation, both in standard and few-shot downstream contexts. Two pre-training objectives dominate the landscape of SSL techniques: Contrastive Learning and Masked Image Modeling. Features (or tokens) extracted from the final transformer attention block -- specifically, the keys, queries, and values -- as well as features obtained after the final block's feed-forward layer, have become a common foundation for addressing downstream tasks. However, in many existing approaches, these pre-trained ViT features are further processed through additional transformation layers, often involving lightweight heads or combined with distillation, to achieve superior task performance. Although such methods can improve task outcomes, to the best of our knowledge, a comprehensive analysis of the intrinsic representation capabilities of unaltered ViT features has yet to be conducted. This study aims to bridge this gap by systematically evaluating the use of these unmodified features across image classification and segmentation tasks, in both standard and few-shot contexts. The classification and segmentation rules that we use are either hyperplane based (as in logistic regression) or cosine-similarity based, both of which rely on the presence of interpretable directions in the ViT's latent space. Based on the previous rules and without the use of additional feature transformations, we conduct an analysis across token types, tasks, and pre-trained ViT models. This study provides insights into the optimal choice for token type and decision rule based on the task, context, and the pre-training objective, while reporting detailed findings on two widely-used datasets.

cs.CV↗

arXiv2025

HAL-NeRF: High Accuracy Localization Leveraging Neural Radiance Fields

Precise camera localization is a critical task in XR applications and robotics. Using only the camera captures as input to a system is an inexpensive option that enables localization in large indoor and outdoor environments, but it presents challenges in achieving high accuracy. Specifically, camera relocalization methods, such as Absolute Pose Regression (APR), can localize cameras with a median translation error of more than $0.5m$ in outdoor scenes. This paper presents HAL-NeRF, a high-accuracy localization method that combines a CNN pose regressor with a refinement module based on a Monte Carlo particle filter. The Nerfacto model, an implementation of Neural Radiance Fields (NeRFs), is used to augment the data for training the pose regressor and to measure photometric loss in the particle filter refinement module. HAL-NeRF leverages Nerfacto's ability to synthesize high-quality novel views, significantly improving the performance of the localization pipeline. HAL-NeRF achieves state-of-the-art results that are conventionally measured as the average of the median per scene errors. The translation error was $0.025m$ and the rotation error was $0.59$ degrees and 0.04m and 0.58 degrees on the 7-Scenes dataset and Cambridge Landmarks datasets respectively, with the trade-off of increased computational time. This work highlights the potential of combining APR with NeRF-based refinement techniques to advance monocular camera relocalization accuracy.

cs.CV↗

arXiv2023

Unsupervised Interpretable Basis Extraction for Concept-Based Visual Explanations

An important line of research attempts to explain CNN image classifier predictions and intermediate layer representations in terms of human-understandable concepts. Previous work supports that deep representations are linearly separable with respect to their concept label, implying that the feature space has directions where intermediate representations may be projected onto, to become more understandable. These directions are called interpretable, and when considered as a set, they may form an interpretable feature space basis. Compared to previous top-down probing approaches which use concept annotations to identify the interpretable directions one at a time, in this work, we take a bottom-up approach, identifying the directions from the structure of the feature space, collectively, without relying on supervision from concept labels. Instead, we learn the directions by optimizing for a sparsity property that holds for any interpretable basis. We experiment with existing popular CNNs and demonstrate the effectiveness of our method in extracting an interpretable basis across network architectures and training datasets. We make extensions to existing basis interpretability metrics and show that intermediate layer representations become more interpretable when transformed with the extracted bases. Finally, we compare the bases extracted with our method with the bases derived with supervision and find that, in one aspect, unsupervised basis extraction has a strength that constitutes a limitation of learning the basis with supervision, and we provide potential directions for future research.

cs.CV↗

arXiv2022

Cross-Stitched Multi-task Dual Recursive Networks for Unified Single Image Deraining and Desnowing

We present the Cross-stitched Multi-task Unified Dual Recursive Network (CMUDRN) model targeting the task of unified deraining and desnowing in a multi-task learning setting. This unified model borrows from the basic Dual Recursive Network (DRN) architecture developed by Cai et al. The proposed model makes use of cross-stitch units that enable multi-task learning across two separate DRN models, each tasked for single image deraining and desnowing, respectively. By fixing cross-stitch units at several layers of basic task-specific DRN networks, we perform multi-task learning over the two separate DRN models. To enable blind image restoration, on top of these structures we employ a simple neural fusion scheme which merges the output of each DRN. The separate task-specific DRN models and the fusion scheme are simultaneously trained by enforcing local and global supervision. Local supervision is applied on the two DRN submodules, and global supervision is applied on the data fusion submodule of the proposed model. Consequently, we both enable feature sharing across task-specific DRN models and control the image restoration behavior of the DRN submodules. An ablation study shows the strength of the hypothesized CMUDRN model, and experiments indicate that its performance is comparable or better than baseline DRN models on the single image deraining and desnowing tasks. Moreover, CMUDRN enables blind image restoration for the two underlying image restoration tasks, by unifying task-specific image restoration pipelines via a naive parametric fusion scheme. The CMUDRN implementation is available at https://github.com/VCL3D/CMUDRN.

cs.CV↗

arXiv2022

Hybrid Skip: A Biologically Inspired Skip Connection for the UNet Architecture

In this work we introduce a biologically inspired long-range skip connection for the UNet architecture that relies on the perceptual illusion of hybrid images, being images that simultaneously encode two images. The fusion of early encoder features with deeper decoder ones allows UNet models to produce finer-grained dense predictions. While proven in segmentation tasks, the network's benefits are down-weighted for dense regression tasks as these long-range skip connections additionally result in texture transfer artifacts. Specifically for depth estimation, this hurts smoothness and introduces false positive edges which are detrimental to the task due to the depth maps' piece-wise smooth nature. The proposed HybridSkip connections show improved performance in balancing the trade-off between edge preservation, and the minimization of texture transfer artifacts that hurt smoothness. This is achieved by the proper and balanced exchange of information that Hybrid-Skip connections offer between the high and low frequency, encoder and decoder features, respectively.

cs.CV↗

arXiv2022

Monocular Spherical Depth Estimation with Explicitly Connected Weak Layout Cues

Spherical cameras capture scenes in a holistic manner and have been used for room layout estimation. Recently, with the availability of appropriate datasets, there has also been progress in depth estimation from a single omnidirectional image. While these two tasks are complementary, few works have been able to explore them in parallel to advance indoor geometric perception, and those that have done so either relied on synthetic data, or used small scale datasets, as few options are available that include both layout annotations and dense depth maps in real scenes. This is partly due to the necessity of manual annotations for room layouts. In this work, we move beyond this limitation and generate a 360 geometric vision (360V) dataset that includes multiple modalities, multi-view stereo data and automatically generated weak layout cues. We also explore an explicit coupling between the two tasks to integrate them into a singleshot trained model. We rely on depth-based layout reconstruction and layout-based depth attention, demonstrating increased performance across both tasks. By using single 360 cameras to scan rooms, the opportunity for facile and quick building-scale 3D scanning arises.

cs.CV↗

arXiv2021

Towards Full-to-Empty Room Generation with Structure-Aware Feature Encoding and Soft Semantic Region-Adaptive Normalization

The task of transforming a furnished room image into a background-only is extremely challenging since it requires making large changes regarding the scene context while still preserving the overall layout and style. In order to acquire photo-realistic and structural consistent background, existing deep learning methods either employ image inpainting approaches or incorporate the learning of the scene layout as an individual task and leverage it later in a not fully differentiable semantic region-adaptive normalization module. To tackle these drawbacks, we treat scene layout generation as a feature linear transformation problem and propose a simple yet effective adjusted fully differentiable soft semantic region-adaptive normalization module (softSEAN) block. We showcase the applicability in diminished reality and depth estimation tasks, where our approach besides the advantages of mitigating training complexity and non-differentiability issues, surpasses the compared methods both quantitatively and qualitatively. Our softSEAN block can be used as a drop-in module for existing discriminative and generative models. Implementation is available on vcl3d.github.io/PanoDR/.

cs.CV↗

arXiv2021

A benchmark with decomposed distribution shifts for 360 monocular depth estimation

In this work we contribute a distribution shift benchmark for a computer vision task; monocular depth estimation. Our differentiation is the decomposition of the wider distribution shift of uncontrolled testing on in-the-wild data, to three distinct distribution shifts. Specifically, we generate data via synthesis and analyze them to produce covariate (color input), prior (depth output) and concept (their relationship) distribution shifts. We also synthesize combinations and show how each one is indeed a different challenge to address, as stacking them produces increased performance drops and cannot be addressed horizontally using standard approaches.

cs.CV↗

arXiv2021

On Coordinate Decoding for Keypoint Estimation Tasks

A series of 2D (and 3D) keypoint estimation tasks are built upon heatmap coordinate representation, i.e. a probability map that allows for learnable and spatially aware encoding and decoding of keypoint coordinates on grids, even allowing for sub-pixel coordinate accuracy. In this report, we aim to reproduce the findings of DARK that investigated the 2D heatmap representation by highlighting the importance of the encoding of the ground truth heatmap and the decoding of the predicted heatmap to keypoint coordinates. The authors claim that a) a more principled distribution-aware coordinate decoding method overcomes the limitations of the standard techniques widely used in the literature, and b), that the reconstruction of heatmaps from ground-truth coordinates by generating accurate and continuous heatmap distributions lead to unbiased model training, contrary to the standard coordinate encoding process that quantizes the keypoint coordinates on the resolution of the input image grid.

cs.CV↗

arXiv2021

HUMAN4D: A Human-Centric Multimodal Dataset for Motions and Immersive Media

We introduce HUMAN4D, a large and multimodal 4D dataset that contains a variety of human activities simultaneously captured by a professional marker-based MoCap, a volumetric capture and an audio recording system. By capturing 2 female and $2$ male professional actors performing various full-body movements and expressions, HUMAN4D provides a diverse set of motions and poses encountered as part of single- and multi-person daily, physical and social activities (jumping, dancing, etc.), along with multi-RGBD (mRGBD), volumetric and audio data. Despite the existence of multi-view color datasets captured with the use of hardware (HW) synchronization, to the best of our knowledge, HUMAN4D is the first and only public resource that provides volumetric depth maps with high synchronization precision due to the use of intra- and inter-sensor HW-SYNC. Moreover, a spatio-temporally aligned scanned and rigged 3D character complements HUMAN4D to enable joint research on time-varying and high-quality dynamic meshes. We provide evaluation baselines by benchmarking HUMAN4D with state-of-the-art human pose estimation and 3D compression methods. For the former, we apply 2D and 3D pose estimation algorithms both on single- and multi-view data cues. For the latter, we benchmark open-source 3D codecs on volumetric data respecting online volumetric video encoding and steady bit-rates. Furthermore, qualitative and quantitative visual comparison between mesh-based volumetric data reconstructed in different qualities showcases the available options with respect to 4D representations. HUMAN4D is introduced to the computer vision and graphics research communities to enable joint research on spatio-temporally aligned pose, volumetric, mRGBD and audio data cues. The dataset and its code are available https://tofis.github.io/myurls/human4d.

cs.CV↗

arXiv2021

DeepMoCap: Deep Optical Motion Capture Using Multiple Depth Sensors and Retro-Reflectors

In this paper, a marker-based, single-person optical motion capture method (DeepMoCap) is proposed using multiple spatio-temporally aligned infrared-depth sensors and retro-reflective straps and patches (reflectors). DeepMoCap explores motion capture by automatically localizing and labeling reflectors on depth images and, subsequently, on 3D space. Introducing a non-parametric representation to encode the temporal correlation among pairs of colorized depthmaps and 3D optical flow frames, a multi-stage Fully Convolutional Network (FCN) architecture is proposed to jointly learn reflector locations and their temporal dependency among sequential frames. The extracted reflector 2D locations are spatially mapped in 3D space, resulting in robust 3D optical data extraction. The subject's motion is efficiently captured by applying a template-based fitting technique on the extracted optical data. Two datasets have been created and made publicly available for evaluation purposes; one comprising multi-view depth and 3D optical flow annotated images (DMC2.5D), and a second, consisting of spatio-temporally aligned multi-view depth images along with skeleton, inertial and ground truth MoCap data (DMC3D). The FCN model outperforms its competitors on the DMC2.5D dataset using 2D Percentage of Correct Keypoints (PCK) metric, while the motion capture outcome is evaluated against RGB-D and inertial data fusion approaches on DMC3D, outperforming the next best method by 4.5% in total 3D PCK accuracy.

cs.CV↗

arXiv2021

Pano3D: A Holistic Benchmark and a Solid Baseline for $360^o$ Depth Estimation

Pano3D is a new benchmark for depth estimation from spherical panoramas. It aims to assess performance across all depth estimation traits, the primary direct depth estimation performance targeting precision and accuracy, and also the secondary traits, boundary preservation, and smoothness. Moreover, Pano3D moves beyond typical intra-dataset evaluation to inter-dataset performance assessment. By disentangling the capacity to generalize to unseen data into different test splits, Pano3D represents a holistic benchmark for $360^o$ depth estimation. We use it as a basis for an extended analysis seeking to offer insights into classical choices for depth estimation. This results in a solid baseline for panoramic depth that follow-up works can build upon to steer future progress.

cs.CV↗

arXiv2021

Hyper360 -- a Next Generation Toolset for Immersive Media

Spherical 360° video is a novel media format, rapidly becoming adopted in media production and consumption of immersive media. Due to its novelty, there is a lack of tools for producing highly engaging interactive 360° video for consumption on a multitude of platforms. In this work, we describe the work done so far in the Hyper360 project on tools for mixed 360° video and 3D content. Furthermore, the first pilots which have been produced with the Hyper360 tools and results of the audience assessment of the produced pilots are presented.

cs.CV↗

arXiv2021

Single-Shot Cuboids: Geodesics-based End-to-end Manhattan Aligned Layout Estimation from Spherical Panoramas

It has been shown that global scene understanding tasks like layout estimation can benefit from wider field of views, and specifically spherical panoramas. While much progress has been made recently, all previous approaches rely on intermediate representations and postprocessing to produce Manhattan-aligned estimates. In this work we show how to estimate full room layouts in a single-shot, eliminating the need for postprocessing. Our work is the first to directly infer Manhattan-aligned outputs. To achieve this, our data-driven model exploits direct coordinate regression and is supervised end-to-end. As a result, we can explicitly add quasi-Manhattan constraints, which set the necessary conditions for a homography-based Manhattan alignment module. Finally, we introduce the geodesic heatmaps and loss and a boundary-aware center of mass calculation that facilitate higher quality keypoint estimation in the spherical domain. Our models and code are publicly available at https://vcl3d.github.io/SingleShotCuboids/.

cs.CV↗

arXiv2021

Serverless Streaming for Emerging Media: Towards 5G Network-Driven Cost Optimization

Immersive 3D media is an emerging type of media that captures, encodes and reconstructs the 3D appearance of people and objects, with applications in tele-presence, teleconference, entertainment, gaming and other fields. In this paper, we discuss a novel concept of live 3D immersive media streaming in a serverless setting. In particular, we present a novel network-centric adaptive streaming framework which deviates from a traditional client-based adaptive streaming used in 2D video. In our framework, the decisions for the production of the transcoding profiles are taken in a centralized manner, by considering consumer metrics vs provisioning costs and inferring an expected consumer quality of experience and behaviour based on them. In addition, we demonstrate that a naive application of the serverless paradigm might be sub optimal under some common immersive 3D media scenarios.

cs.NI↗