Alessio Tonioni

Alessio Tonioni

I’m a researcher in computer vision and deep learning and I’m currently working with Federico Tombari at Google Zurich. Previously I was enrolled as a post doc at the Computer Vision Lab of the university of Bologna under the supervision of Professor Luigi Di Stefano. I received my PhD in Computer Science and Engineering from University of Bologna on April 2019. During my PhD I have worked on deep learning solutions for product detection and recognition in retail environments and on deep learning applied to depth estimation from stereo and monocular cameras.
Authored Publications
Sort By
  • Title
  • Title, descending
  • Year
  • Year, descending
Editing Everything Everywhere All at Once
Fabio Quattrini
Carmine Zaccagnino
Enis Simsar
Marta Tintore Gazulla
Rita Cucchiara
Silvia Cascianelli
2026
Preview abstract Editing multiple elements of an image in a single forward pass has recently emerged as a practical alternative to multi-turn image manipulation, offering improved efficiency. However, when several instructions target different regions, semantic interference often leads to attribute leakage and poor edit disentanglement, especially as the number of edits increases. In this work, we propose MICE (Multi-Instance Concurrent Editing), a training-free strategy for scalable multi-instance image editing with Multimodal Diffusion Transformers. MICE modifies the additive bias of joint attention to regulate interactions between instance-specific text, latent, and context tokens identified via user-provided segmentation masks. Specifically, MICE allows intra-instance attention, penalizes interactions between neighboring region tokens, and suppresses unrelated cross-instance attention. As a result, our method enforces attribute binding while preserving global visual consistency. We evaluate MICE on LoMOE-Bench and introduce MICE-Bench, a more challenging benchmark with an average of 8.5 concurrent edits per image. The experiments demonstrate that our approach outperforms strong baselines and recent competitors in terms of the number of attempted edits and faithfulness to the textual editing instruction. View details
RefAM: Attention Magnets for Zero-Shot Referral Segmentation
Anna Kukleva
Enis Simsar
Ferjad Naeem
Jan Eric Lenssen
Bernt Schiele
2026
Preview abstract Internal features from large-scale pre-trained diffusion models are known to encode rich semantic information. However, standard approaches extract features by feeding noisy images into the model, degrading the quality of the representation. We propose InvertDIFT, a method that leverages diffusion inversion to recover clean representations without modifying the model. By inverting the noisy forward process to retrieve clean or intermediate states, we extract timestep-independent features from the diffusion model as is. This approach preserves the model's pre-trained knowledge and eliminates the need for additional training, while achieving performance on par or better than fine-tuned approaches across several tasks. View details
Preview abstract Reinforcement learning (RL) with verifiable rewards (RLVR) has demonstrated the great potential of enhancing the reasoning abilities in multimodal large language models (MLLMs). However, the reliance on language-centric priors and expensive manual annotations prevents MLLMs' intrinsic visual understanding and scalable reward designs. In this work, we introduce SSL-R1, a generic self-supervised RL framework that derives verifiable rewards directly from images. To this end, we revisit self-supervised learning (SSL) in visual domains and reformulate widely-used SSL tasks into a set of verifiable visual puzzles for RL-based post-training, requiring neither human nor external model supervision. Training MLLMs on these tasks substantially improves their performance on multimodal understanding and reasoning benchmarks, highlighting the potential of leveraging self-supervised vision-centric tasks for MLLM post-training. We hope this work will provide useful experience in devising effective self-supervised verifiable rewards to enable RLVR at scale. View details
Autoregressive styled text image generation, but make it reliable
Carmini Zaccagnino
Fabio Quattrini
Vittorio Pippi
Silvia Cascianelli
Rita Cucchiara
2026
Preview abstract The task of generating images of styled text (often referred to as Styled Handwritten Text Generation - HTG) is attracting attention from different communities, including Computer Vision, Document Analysis, and Graphic Design. A lot of research effort in this task is dedicated to developing strategies that reproduce the stylistic characteristics of a given writer, with impressive results in terms of adherence and generalization achieved by the recently proposed Autoregressive Transformer paradigm for HGT. However, faithfulness to the desired text to render is not yet guaranteed, often due to collapse, generation artifacts, and lack of controllability in autoreressive models. In this work, we extend the autoregressive formulation by framing HTG as a multimodal prompt-conditioned generation task and tackle the content controllability issues by introducing text input tokens for better alignment with the visual ones and by devising a Classifier-Free-Guidance-based training strategy for our autoregressive model. Through extensive experimental validation, we demonstrate that our approach, dubbed Eruku, not only can generalize well to unseen styles but can also better follow the textual prompt to achieve content faithfulness. View details
MOSAIC-GS: MOnocular Scene Reconstruction via Advanced Initialization for Complex Dynamic Environments
Svitlana Morkva
Max Wilder-Smith
Michael Oechsle
Marco Hutter
Vaishakh Patil
2026
Preview abstract We present MOSAIC-GS, a novel, fully explicit, and computationally efficient approach for high-fidelity dynamic scene reconstruction from monocular videos. Monocular reconstruction is inherently ill-posed due to the absence of sufficient multiview constraints, making accurate recovery of object geometry and temporal coherence particularly challenging. To address this, we leverage multiple geometric cues, such as depth, optical flow, dynamic object segmentation, and tracking trajectories, combined with rigidity constraints to estimate preliminary 3D scene dynamics during an advanced initialization stage. Recovering scene flow prior to the photometric optimization phase reduces reliance on motion inference from visual appearance alone, which is often ambiguous in monocular settings. To enable compact representations, fast training, and real-time rendering while supporting non-rigid deformations, the scene is decomposed into static and dynamic components, with dynamic trajectories represented as time-dependent Poly-Fourier curves for parameter-efficient motion encoding. We demonstrate that MOSAIC-GS achieves substantially faster optimization and rendering compared to existing methods, while maintaining reconstruction quality on par with state-of-the-art approaches across standard monocular dynamic scene benchmarks. View details
Shifting the Breaking Point of Flow Matching for Multi-Instance Editing
Carmine Zaccagnino
Fabio Quattrini
Enis Simsar
Marta Tintore Gazulla
Rita Cucchiara
Silvia Cascianelli
2026
Preview abstract Flow matching models have recently emerged as an efficient alternative to diffusion, especially for text-guided image generation and editing, offering faster inference through continuous-time dynamics. However, existing flow-based editors predominantly support global or single-instruction edits and struggle with multi-instance scenarios, where multiple parts of a reference input must be edited independently without semantic interference. We identify this limitation as a consequence of globally conditioned velocity fields and joint attention mechanisms, which entangle concurrent edits. To address this issue, we introduce Instance-Disentangled Attention, a mechanism that partitions joint attention operations, enforcing binding between instance-specific textual instructions and spatial regions during velocity field estimation. We evaluate our approach on both natural image editing and a newly introduced benchmark of text-dense infographics with region-level editing instructions. Experimental results demonstrate that our approach promotes edit disentanglement and locality while preserving global output coherence, enabling single-pass, instance-level editing. View details
DataComp for Vision-Language Models
Matteo Farina
Vishaal Udandarao
Thao Nguyen
Selim Kuzucu
Maximilian Böther
Andreas Hochlehnert
Adhiraj Ghosh
Marianna Nezhurina
Karsten Roth
Joschka Struber
Sebastian Dziadzio
Yuhui Zhang
Elaine Sui
Soumya Jahagirdhar
Dhruba Ghosh
Hasan Hammoud
Thomas De Min
Simone Caldarella
Jehanzeb Mirza
Sedrick Keh
Mehdi Cherti
Hilde Kuehne
Bernt Schiele
Serena Yeung
Ferjad Naeem
Ana Klimovic
Elisa Ricci
Matthias Bethge
Sewoong Oh
Ameya Prabhu
Jenia Jitsev
Massimiliano Mancini
Ludwig Schmidt
Nikhil Parthasarathy
(2026)
Preview abstract Building performant Vision-Language Models (VLMs) requires carefully curating large-scale training datasets, yet the community lacks systematic benchmarks for evaluating such curation strategies. We introduce DataComp for VLMs (DCVLM), a benchmark for controlled data-centric experiments to improve VLM training. As part of DCVLM, we collect 160 datasets spanning four data types -- image-caption pairs, multimodal interleaved documents, text-only, and instruction-tuning data -- into a corpus of 6T multimodal tokens. DCVLM allows participants to test curation strategies (filtering, mixing, formatting, sampling) across 1B-8B models and 6.25B-200B token budgets. Models are then evaluated on a carefully selected suite of up to 52 downstream benchmarks across 9 domains. We conduct extensive experiments on DCVLM and find that data mixing, not filtering, is key to a high-quality training dataset: instruction-heavy mixtures scale better than caption-heavy ones, with gains widening at larger scales. The resulting dataset, DCVLM-Baseline, enables training an 8B VLM to 63.6% accuracy on our 33-task core suite with 200B training tokens. Compared to FineVision, the state-of-the-art open VLM training dataset, this represents an improvement of +5.4pp. DCVLM and all accompanying artifacts will be made publicly available at this https URL. View details
Snap-it, Tap-it, Splat-it: Tactile-Informed 3D Gaussian Splatting for Reconstructing Challenging Surfaces
Mauro Comi
Max Yang
Jonathan Tremblay
Valts Blukis
Yijiong Lin
Nathan Lepora
Laurence Aitchison
2025
Preview abstract Touch and vision go hand in hand, mutually enhancing our ability to understand the world. From a research perspective, the problem of mixing touch and vision is underexplored and presents interesting challenges. To this end, we propose Tactile-Informed 3DGS, a novel approach that incorporates touch data (local depth maps) with multi-view vision data to achieve surface reconstruction and novel view synthesis. Our method optimises 3D Gaussian primitives to accurately model the object's geometry at points of contact. By creating a framework that decreases the transmittance at touch locations, we achieve a refined surface reconstruction, ensuring a uniformly smooth depth map. Touch is particularly useful when considering non-Lambertian objects (e.g. shiny or reflective surfaces) since contemporary methods tend to fail to reconstruct with fidelity specular highlights. By combining vision and tactile sensing, we achieve more accurate geometry reconstructions with fewer images than prior methods. We conduct evaluation on objects with glossy and reflective surfaces and demonstrate the effectiveness of our approach, offering significant improvements in reconstruction quality. View details
Preview abstract Understanding fine-grained temporal dynamics is crucial in egocentric videos, where continuous streams capture frequent, close-up interactions with objects. In this work, we bring to light that current egocentric video question-answering datasets often include questions that can be answered using only few frames or commonsense reasoning, without being necessarily grounded in the actual video. Our analysis shows that state-of-the-art Multi-Modal Large Language Models (MLLMs) on these benchmarks achieve remarkably high performance using just text or a single frame as input. To address these limitations, we introduce EgoTempo, a dataset specifically designed to evaluate temporal understanding in the egocentric domain. EgoTempo emphasizes tasks that require integrating information across the entire video, ensuring that models would need to rely on temporal patterns rather than static cues or pre-existing knowledge. Extensive experiments on EgoTempo show that current MLLMs still fall short in temporal reasoning on egocentric videos, and thus we hope EgoTempo will catalyze new research in the field and inspire models that better capture the complexity of temporal dynamics. Dataset and code are available at https://github.com/google-research-datasets/egotempo.git. View details
Training-free Online Video Step Grounding
Luca Zanella
Massimiliano Mancini
Yiming Wang
Elisa Ricci
2025
Preview abstract Given a task and a set of steps composing it, video step grounding aims to detect which steps are performed in a video. Standard approaches for this task require a labeled training set with step-level annotations. This is costly obtain and may suffer from dataset imbalance for rare tasks. Moreover, the task is performed offline, i.e., processing the full video at once. This is a disadvantage, especially for environments requiring online decisions. In this work, we explore how to perform this task \textit{online} and \textit{without training}. We achieve this by using recent large multimodal models (LMMs), showing how their zero-shot capabilities are sufficient to outperform training-based models. We then propose a filtering mechanism that refines LMMs predictions by (i) adding a verification step on the model's output and (ii) modeling the step-transition probabilities via a dependency matrix and a transition one, both obtained through large language models. Experiments on three dataset show the efficacy of the approach. View details
×