arXiv subjects

Riccardo Renzulli

Publications and source records attributed to Riccardo Renzulli.

17 recordsLinked to original sources

arXiv2026

TORINO: Token Reduction via Interpretable Concept Overlap in Vision-Language Models

Vision-Language Models (VLMs) have demonstrated impressive capabilities across different tasks, but their computational cost is dominated by the large number of visual tokens fed to the language model. Existing token reduction methods rely on attention-based scores or pairwise similarity, without an explicit semantic representation of each token. We introduce TORINO (TOken Reduction via Interpretable coNcept Overlap), a plug-and-play framework for adaptive visual token reduction in VLMs that requires no fine-tuning of the underlying model. TORINO leverages Sparse Autoencoders (SAEs) to project visual tokens into an interpretable latent space where token relationships can be analyzed through shared concept activations. Specifically, we define concept overlap as the degree of agreement between active SAE latents and use it to group tokens that share semantic content. Reduction within each group is then performed by either pruning or merging, providing a unified framework that preserves semantically important visual information while removing redundancy. Unlike fixed-budget approaches, TORINO dynamically adapts the reduction rate to input complexity, allowing different images to retain different numbers of tokens. Experiments across multiple vision-language benchmarks show that TORINO achieves favorable efficiency-accuracy trade-offs, reducing the number of visual tokens with minimal performance loss.

cs.CV↗

arXiv2026

Look But Don't Touch with Sparse Autoencoders for Unlearning in Diffusion Models

Sparse autoencoders (SAEs) have recently been proposed as interpretable tools for concept-level manipulation, under the assumption that isolated features can serve as controllable intervention points. In this work, we systematically evaluate this assumption in the context of object erasure and steering in diffusion models. We show that while SAEs reliably detect and localize semantic concepts within diffusion model activations, direct intervention in their latent space frequently induces out-of-distribution activations, resulting in severe visual artifacts. To disentangle detection from intervention, we use SAE activations purely as semantic detectors to identify image regions containing the target object, and replace those patch embeddings with the ones that do not contain it. This detection-based replacement preserves the diffusion model's activation statistics and produces significantly cleaner erasure results than latent steering. Our findings reveal a fundamental gap between concept detection and concept intervention in diffusion models: monosemantic or sparse features are not inherently suitable as control knobs for steering. These results position SAEs as powerful interpretability tools for analyzing generative models, but highlight important limitations when used for direct manipulation, such as unlearning.

cs.CV↗

arXiv2026

SAEmnesia: Erasing Concepts in Diffusion Models with Supervised Sparse Autoencoders

Concept unlearning in diffusion models is hampered by feature splitting, where concepts are distributed across many latent features, making their removal challenging and computationally expensive. We introduce SAEmnesia, a supervised sparse autoencoder framework that overcomes this by enforcing one-to-one concept-neuron mappings. By systematically labeling concepts during training, our method achieves feature centralization, binding each concept to a single, interpretable neuron. This enables highly targeted and efficient concept erasure. Compared to the state-of-the-art sparse autoencoder-based unlearning approach, SAEmnesia reduces hyperparameter search by 96.67% and achieves a 9.22% improvement on the UnlearnCanvas benchmark for objects. Our method also shows superior scalability in sequential unlearning, improving accuracy by 28.4% when removing nine objects, establishing a step forward for precise and controllable concept erasure. Moreover, SAEmnesia effectively suppresses nudity on the I2P benchmark and remains robust to adversarial attacks. Source code available at https://github.com/EIDOSLAB/SAEmnesia.

cs.CV↗

arXiv2026

MedSAE: Dissecting MedCLIP Representations with Sparse Autoencoders

Artificial intelligence in healthcare requires models that are accurate and interpretable. We advance mechanistic interpretability in medical vision by applying Medical Sparse Autoencoders (MedSAEs) to the latent space of MedCLIP, a vision-language model trained on chest radiographs and reports. To quantify interpretability, we propose an evaluation framework that combines correlation metrics, entropy analyses, and automated neuron naming via the MedGemma foundation model. Experiments on the CheXpert dataset show that MedSAE neurons achieve higher monosemanticity and interpretability than raw MedCLIP features. Our findings bridge high-performing medical AI and transparency, offering a scalable step toward clinically reliable representations. The source code supporting the findings of this study is available at https://github.com/EIDOSLAB/MedSAE.

cs.AI↗

arXiv2026

GoDe: Gaussians on Demand for Progressive Level of Detail and Scalable Compression

Recent progress in compressing explicit radiance field representations, particularly 3D Gaussian Splatting, has substantially reduced memory consumption while improving real-time rendering performance. However, existing approaches remain inherently single-rate: each compression level requires a separately optimized model, yielding a set of fixed operating points rather than a truly scalable representation. This limits deployment in scenarios where memory, bandwidth, or computational budgets vary across devices or over time. We argue that scalability should be an intrinsic property of the representation. We show that trained explicit radiance models exhibit a structured distribution of information, which can be revealed using standard optimization signals available during training. In particular, aggregated gradient sensitivity provides a simple, model-agnostic criterion to organize primitives from coarse structure to finer refinements. Building on this, we introduce GoDe (Gaussians on Demand), a general framework for scalable compression and progressive level-of-detail control, instantiated for 3D Gaussian Splatting. Starting from a single trained model, GoDe reorganizes Gaussian primitives into a fixed progressive hierarchy supporting multiple rate-distortion operating points without retraining or per-level fine-tuning. A single quantization-aware fine-tuning stage ensures consistent behavior across all levels under low-precision storage. Extensive experiments on standard benchmarks and multiple 3D Gaussian Splatting backbones show that GoDe achieves rate-distortion performance comparable to state-of-the-art single-rate methods, while enabling truly scalable compression and adaptive rendering within a unified representation. Project page: https://gaussians-on-demand.github.io

cs.CV↗

arXiv2026

Automated Prediction of Paravalvular Regurgitation before Transcatheter Aortic Valve Implantation

Severe aortic stenosis is a common and life-threatening condition in elderly patients, often treated with Transcatheter Aortic Valve Implantation (TAVI). Despite procedural advances, paravalvular aortic regurgitation (PVR) remains one of the most frequent post-TAVI complications, with a proven impact on long-term prognosis. In this work, we investigate the potential of deep learning to predict the occurrence of PVR from preoperative cardiac CT. To this end, a dataset of preoperative TAVI patients was collected, and 3D convolutional neural networks were trained on isotropic CT volumes. The results achieved suggest that volumetric deep learning can capture subtle anatomical features from pre-TAVI imaging, opening new perspectives for personalized risk assessment and procedural optimization. Source code is available at https://github.com/EIDOSLAB/tavi.

cs.CV↗

arXiv2026

Synthetic Dataset Generation and Validation for Robotic Surgery Instrument Segmentation

This paper presents a comprehensive workflow for generating and validating a synthetic dataset designed for robotic surgery instrument segmentation. A 3D reconstruction of the Da Vinci robotic arms was refined and animated in Autodesk Maya through a fully automated Python-based pipeline capable of producing photorealistic, labeled video sequences. Each scene integrates randomized motion patterns, lighting variations, and synthetic blood textures to mimic intraoperative variability while preserving pixel-accurate ground truth masks. To validate the realism and effectiveness of the generated data, several segmentation models were trained under controlled ratios of real and synthetic data. Results demonstrate that a balanced composition of real and synthetic samples significantly improves model generalization compared to training on real data only, while excessive reliance on synthetic data introduces a measurable domain shift. The proposed framework provides a reproducible and scalable tool for surgical computer vision, supporting future research in data augmentation, domain adaptation, and simulation-based pretraining for robotic-assisted surgery. Data and code are available at https://github.com/EIDOSLAB/Sintetic-dataset-DaVinci.

cs.CV↗

arXiv2026

Cardiac Output Prediction from Echocardiograms: Self-Supervised Learning with Limited Data

Cardiac Output (CO) is a key parameter in the diagnosis and management of cardiovascular diseases. However, its accurate measurement requires right-heart catheterization, an invasive and time-consuming procedure, motivating the development of reliable non-invasive alternatives using echocardiography. In this work, we propose a self-supervised learning (SSL) pretraining strategy based on SimCLR to improve CO prediction from apical four-chamber echocardiographic videos. The pretraining is performed using the same limited dataset available for the downstream task, demonstrating the potential of SSL even under data scarcity. Our results show that SSL mitigates overfitting and improves representation learning, achieving an average Pearson correlation of 0.41 on the test set and outperforming PanEcho, a model trained on over one million echocardiographic exams. Source code is available at https://github.com/EIDOSLAB/cardiac-output.

cs.CV↗

arXiv2025

Capsule Networks Do Not Need to Model Everything

Capsule networks are biologically inspired neural networks that group neurons into vectors called capsules, each explicitly representing an object or one of its parts. The routing mechanism connects capsules in consecutive layers, forming a hierarchical structure between parts and objects, also known as a parse tree. Capsule networks often attempt to model all elements in an image, requiring large network sizes to handle complexities such as intricate backgrounds or irrelevant objects. However, this comprehensive modeling leads to increased parameter counts and computational inefficiencies. Our goal is to enable capsule networks to focus only on the object of interest, reducing the number of parse trees. We accomplish this with REM (Routing Entropy Minimization), a technique that minimizes the entropy of the parse tree-like structure. REM drives the model parameters distribution towards low entropy configurations through a pruning mechanism, significantly reducing the generation of intra-class parse trees. This empowers capsules to learn more stable and succinct representations with fewer parameters and negligible performance loss.

cs.CV↗

arXiv2025

When Does Pruning Benefit Vision Representations?

Pruning is widely used to reduce the complexity of deep learning models, but its effects on interpretability and representation learning remain poorly understood. This paper investigates how pruning influences vision models across three key dimensions: (i) interpretability, (ii) unsupervised object discovery, and (iii) alignment with human perception. We first analyze different vision network architectures to examine how varying sparsity levels affect feature attribution interpretability methods. Additionally, we explore whether pruning promotes more succinct and structured representations, potentially improving unsupervised object discovery by discarding redundant information while preserving essential features. Finally, we assess whether pruning enhances the alignment between model representations and human perception, investigating whether sparser models focus on more discriminative features similarly to humans. Our findings also reveal the presence of sweet spots, where sparse models exhibit higher interpretability, downstream generalization and human alignment. However, these spots highly depend on the network architectures and their size in terms of trainable parameters. Our results suggest a complex interplay between these three dimensions, highlighting the importance of investigating when and how pruning benefits vision representations.

cs.CV↗

arXiv2025

Unsupervised contrastive analysis for anomaly detection in brain MRIs via conditional diffusion models

Contrastive Analysis (CA) detects anomalies by contrasting patterns unique to a target group (e.g., unhealthy subjects) from those in a background group (e.g., healthy subjects). In the context of brain MRIs, existing CA approaches rely on supervised contrastive learning or variational autoencoders (VAEs) using both healthy and unhealthy data, but such reliance on target samples is challenging in clinical settings. Unsupervised Anomaly Detection (UAD) offers an alternative by learning a reference representation of healthy anatomy without the need for target samples. Deviations from this reference distribution can indicate potential anomalies. In this context, diffusion models have been increasingly adopted in UAD due to their superior performance in image generation compared to VAEs. Nonetheless, precisely reconstructing the anatomy of the brain remains a challenge. In this work, we propose an unsupervised framework to improve the reconstruction quality by training a self-supervised contrastive encoder on healthy images to extract meaningful anatomical features. These features are used to condition a diffusion model to reconstruct the healthy appearance of a given image, enabling interpretable anomaly localization via pixel-wise comparison. We validate our approach through a proof-of-concept on a facial image dataset and further demonstrate its effectiveness on four brain MRI datasets, achieving state-of-the-art anomaly localization performance on the NOVA benchmark.

cs.CV↗

arXiv2024

Boost Your NeRF: A Model-Agnostic Mixture of Experts Framework for High Quality and Efficient Rendering

Since the introduction of NeRFs, considerable attention has been focused on improving their training and inference times, leading to the development of Fast-NeRFs models. Despite demonstrating impressive rendering speed and quality, the rapid convergence of such models poses challenges for further improving reconstruction quality. Common strategies to improve rendering quality involves augmenting model parameters or increasing the number of sampled points. However, these computationally intensive approaches encounter limitations in achieving significant quality enhancements. This study introduces a model-agnostic framework inspired by Sparsely-Gated Mixture of Experts to enhance rendering quality without escalating computational complexity. Our approach enables specialization in rendering different scene components by employing a mixture of experts with varying resolutions. We present a novel gate formulation designed to maximize expert capabilities and propose a resolution-based routing technique to effectively induce sparsity and decompose scenes. Our work significantly improves reconstruction quality while maintaining competitive performance.

cs.CV↗

arXiv2024

Hierarchical Object-Centric Learning with Capsule Networks

Capsule networks (CapsNets) were introduced to address convolutional neural networks limitations, learning object-centric representations that are more robust, pose-aware, and interpretable. They organize neurons into groups called capsules, where each capsule encodes the instantiation parameters of an object or one of its parts. Moreover, a routing algorithm connects capsules in different layers, thereby capturing hierarchical part-whole relationships in the data. This thesis investigates the intriguing aspects of CapsNets and focuses on three key questions to unlock their full potential. First, we explore the effectiveness of the routing algorithm, particularly in small-sized networks. We propose a novel method that anneals the number of routing iterations during training, enhancing performance in architectures with fewer parameters. Secondly, we investigate methods to extract more effective first-layer capsules, also known as primary capsules. By exploiting pruned backbones, we aim to improve computational efficiency by reducing the number of capsules while achieving high generalization. This approach reduces CapsNets memory requirements and computational effort. Third, we explore part-relationship learning in CapsNets. Through extensive research, we demonstrate that capsules with low entropy can extract more concise and discriminative part-whole relationships compared to traditional capsule networks, even with reasonable network sizes. Lastly, we showcase how CapsNets can be utilized in real-world applications, including autonomous localization of unmanned aerial vehicles, quaternion-based rotations prediction in synthetic datasets, and lung nodule segmentation in biomedical imaging. The findings presented in this thesis contribute to a deeper understanding of CapsNets and highlight their potential to address complex computer vision challenges.

cs.CV↗

arXiv2024

AI-Assisted Diagnosis for Covid-19 CXR Screening: From Data Collection to Clinical Validation

In this paper, we present the major results from the Covid Radiographic imaging System based on AI (Co.R.S.A.) project, which took place in Italy. This project aims to develop a state-of-the-art AI-based system for diagnosing Covid-19 pneumonia from Chest X-ray (CXR) images. The contributions of this work are manyfold: the release of the public CORDA dataset, a deep learning pipeline for Covid-19 detection, and the clinical validation of the developed solution by expert radiologists. The proposed detection model is based on a two-step approach that, paired with state-of-the-art debiasing, provides reliable results. Most importantly, our investigation includes the actual usage of the diagnosis aid tool by radiologists, allowing us to assess the real benefits in terms of accuracy and time efficiency. Project homepage: https://corsa.di.unito.it/

eess.IV↗

arXiv2022

LSVL: Large-scale season-invariant visual localization for UAVs

Localization of autonomous unmanned aerial vehicles (UAVs) relies heavily on Global Navigation Satellite Systems (GNSS), which are susceptible to interference. Especially in security applications, robust localization algorithms independent of GNSS are needed to provide dependable operations of autonomous UAVs also in interfered conditions. Typical non-GNSS visual localization approaches rely on known starting pose, work only on a small-sized map, or require known flight paths before a mission starts. We consider the problem of localization with no information on initial pose or planned flight path. We propose a solution for global visual localization on a map at scale up to 100 km2, based on matching orthoprojected UAV images to satellite imagery using learned season-invariant descriptors. We show that the method is able to determine heading, latitude and longitude of the UAV at 12.6-18.7 m lateral translation error in as few as 23.2-44.4 updates from an uninformed initialization, also in situations of significant seasonal appearance difference (winter-summer) between the UAV image and the map. We evaluate the characteristics of multiple neural network architectures for generating the descriptors, and likelihood estimation methods that are able to provide fast convergence and low localization error. We also evaluate the operation of the algorithm using real UAV data and evaluate running time on a real-time embedded platform. We believe this is the first work that is able to recover the pose of an UAV at this scale and rate of convergence, while allowing significant seasonal difference between camera observations and map.

cs.RO↗

arXiv2022

Towards Efficient Capsule Networks

From the moment Neural Networks dominated the scene for image processing, the computational complexity needed to solve the targeted tasks skyrocketed: against such an unsustainable trend, many strategies have been developed, ambitiously targeting performance's preservation. Promoting sparse topologies, for example, allows the deployment of deep neural networks models on embedded, resource-constrained devices. Recently, Capsule Networks were introduced to enhance explainability of a model, where each capsule is an explicit representation of an object or its parts. These models show promising results on toy datasets, but their low scalability prevents deployment on more complex tasks. In this work, we explore sparsity besides capsule representations to improve their computational efficiency by reducing the number of capsules. We show how pruning with Capsule Network achieves high generalization with less memory requirements, computational effort, and inference and training time.

cs.CV↗

arXiv2022

Lung nodules segmentation from CT with DeepHealth toolkit

The accurate and consistent border segmentation plays an important role in the tumor volume estimation and its treatment in the field of Medical Image Segmentation. Globally, Lung cancer is one of the leading causes of death and the early detection of lung nodules is essential for the early cancer diagnosis and survival rate of patients. The goal of this study was to demonstrate the feasibility of Deephealth toolkit including PyECVL and PyEDDL libraries to precisely segment lung nodules. Experiments for lung nodules segmentation has been carried out on UniToChest using PyECVL and PyEDDL, for data pre-processing as well as neural network training. The results depict accurate segmentation of lung nodules across a wide diameter range and better accuracy over a traditional detection approach. The datasets and the code used in this paper are publicly available as a baseline reference.

eess.IV↗