arXiv subjects

Matteo Dunnhofer

Publications and source records attributed to Matteo Dunnhofer.

At least 19 recordsLinked to original sources

arXiv2026

Primate vision reveals a missing principle for robust dynamic AI

How does an intelligent visual system combine what objects look like with how they move while remaining robust as appearance changes? We addressed this question by comparing human perception and neural activity in macaque inferior temporal cortex with representations from image- and video-based neural networks spanning recognition, segmentation, optic-flow processing and predictive world modeling. Temporal integration improved object representations, but most video recognition models generalized poorly when appearance was disrupted while motion structure was preserved. Humans and macaque IT remained robust. Notably, predictive world models combined strong cross-appearance generalization with the closest correspondence to IT, outperforming other video-modeling approaches in neural fidelity. Yet no model reproduced the cortical transformation from early appearance-dominated responses toward later appearance-invariant motion coding. These results identify progressive integration of motion into object representations as a principle of robust dynamic vision and implicate predictive learning as a promising route toward realizing this computation in artificial systems.

cs.CV↗

arXiv2026

CPUBone: Efficient Vision Backbone Design for Devices with Low Parallelization Capabilities

Recent research on vision backbone architectures has predominantly focused on optimizing efficiency for hardware platforms with high parallel processing capabilities. This category increasingly includes embedded systems such as mobile phones and embedded AI accelerator modules. In contrast, CPUs do not have the possibility to parallelize operations in the same manner, wherefore models benefit from a specific design philosophy that balances amount of operations (MACs) and hardware-efficient execution by having high MACs per second (MACpS). In pursuit of this, we investigate two modifications to standard convolutions, aimed at reducing computational cost: grouping convolutions and reducing kernel sizes. While both adaptations substantially decrease the total number of MACs required for inference, sustaining low latency necessitates preserving hardware-efficiency. Our experiments across diverse CPU devices confirm that these adaptations successfully retain high hardware-efficiency on CPUs. Based on these insights, we introduce CPUBone, a new family of vision backbone models optimized for CPU-based inference. CPUBone achieves state-of-the-art Speed-Accuracy Trade-offs (SATs) across a wide range of CPU devices and effectively transfers its efficiency to downstream tasks such as object detection and semantic segmentation. Models and code are available at https://github.com/altair199797/CPUBone.

cs.CV↗

arXiv2026

Beyond MACs: Hardware Efficient Architecture Design for Vision Backbones

Vision backbone networks play a central role in modern computer vision. Enhancing their efficiency directly benefits a wide range of downstream applications. To measure efficiency, many publications rely on MACs (Multiply Accumulate operations) as a predictor of execution time. In this paper, we experimentally demonstrate the shortcomings of such a metric, especially in the context of edge devices. By contrasting the MAC count and execution time of common architectural design elements, we identify key factors for efficient execution and provide insights to optimize backbone design. Based on these insights, we present LowFormer, a novel vision backbone family. LowFormer features a streamlined macro and micro design that includes Lowtention, a lightweight alternative to Multi-Head Self-Attention. Lowtention not only proves more efficient, but also enables superior results on ImageNet. Additionally, we present an edge GPU version of LowFormer, that can further improve upon its baseline's speed on edge GPU and desktop GPU. We demonstrate LowFormer's wide applicability by evaluating it on smaller image classification datasets, as well as adapting it to several downstream tasks, such as object detection, semantic segmentation, image retrieval, and visual object tracking. LowFormer models consistently achieve remarkable speed-ups across various hardware platforms compared to recent state-of-the-art backbones. Code and models are available at https://github.com/altair199797/LowFormer/blob/main/Beyond_MACs.md.

cs.CV↗

arXiv2026

The macaque IT cortex but not current artificial vision networks encode object position in perceptually aligned coordinates

Efficient interaction with the visual world requires not only accurate object identification but also precise localization of objects in space. While spatial ("where") processing has traditionally been attributed to dorsal stream pathways, recent work has shown that object position can also be decoded from responses in ventral stream areas such as the inferior temporal (IT) cortex. However, because object position in these paradigms is tightly coupled to pixel-based location, it remains unclear whether ventral stream position signals reflect perceptually meaningful spatial representations or simply inherited retinotopic structure. To address this question, we used the motion aftereffect, a classic visual illusion that shifts perceived object position without changing retinal input. Combining large-scale intracortical recordings in macaque IT with matched human psychophysics, we found that motion adaptation induces systematic direction-opponent biases in IT population codes for object position that mirror human perceptual reports, despite identical pixel-level stimuli. These effects are accompanied by adaptation-driven changes in the geometry of IT population representations. We further tested whether artificial vision systems exhibit similar dynamics. Standard feedforward, recurrent, and state-of-the-art video-based neural networks accurately encode object position but fail to produce adaptation-induced position shifts. However, applying empirically derived transformations based on IT adaptation dynamics to model feature spaces is sufficient to generate similar biases. Together, these results indicate that IT represents object position in perceptually aligned coordinates and also highlight a gap between biological and artificial vision systems in capturing history-dependent spatial coding.

q-bio.NC↗

arXiv2026

Modeling Dynamic Computations in the Primate Ventral Visual Stream

A major goal of computational neuroscience has been to explain how the primate ventral visual stream (VVS) transforms visual input into temporally evolving neural representations that support robust visual perception. Historically, most modeling efforts have assumed static conditions: monkeys fixate a dot, images are briefly flashed, and neural responses are analyzed through time-averaged metrics. Feedforward deep networks trained on static object recognition tasks outperform prior work in approximating these static snapshot-driven VVS responses. However, mounting neurophysiological evidence demonstrates that VVS responses are rich dynamical signals shaped not only by the retinal input but also by intrinsic circuit dynamics, recurrent interactions, and widespread top-down modulation. Moreover, real-world vision is inherently dynamic: objects move, the observer moves, and the eyes actively sample the environment. Here, we review recent progress in modeling dynamic responses in the macaque ventral stream across three domains: (1) intrinsic dynamics elicited by static images, (2) dynamics evoked by dynamic visual stimuli, and (3) dynamics generated by active sensing during eye movements. We argue that accurately modeling VVS dynamics will require representational, circuit-level, and behavioral perspectives, including multi-area recurrence, structured E/I interactions, and temporal objectives that better reflect natural behavior. We outline some key missing ingredients and propose a roadmap toward dynamic, multi-timescale models of the primate VVS.

q-bio.NC↗

arXiv2026

Better, But Not Sufficient: Testing Video ANNs Against Macaque IT Dynamics

Feedforward artificial neural networks (ANNs) trained on static images remain the dominant models of the the primate ventral visual stream, yet they are intrinsically limited to static computations. The primate world is dynamic, and the macaque ventral visual pathways, specifically the inferior temporal (IT) cortex not only supports object recognition but also encodes object motion velocity during naturalistic video viewing. Does IT's temporal responses reflect nothing more than time-unfolded feedforward transformations, framewise features with shallow temporal pooling, or do they embody richer dynamic computations? We tested this by comparing macaque IT responses during naturalistic videos against static, recurrent, and video-based ANN models. Video models provided modest improvements in neural predictivity, particularly at later response stages, raising the question of what kind of dynamics they capture. To probe this, we applied a stress test: decoders trained on naturalistic videos were evaluated on "appearance-free" variants that preserve motion but remove shape and texture. IT population activity generalized across this manipulation, but all ANN classes failed. Thus, current video models better capture appearance-bound dynamics rather than the appearance-invariant temporal computations expressed in IT, underscoring the need for new objectives that encode biological temporal statistics and invariances.

cs.CV↗

arXiv2025

Is Tracking really more challenging in First Person Egocentric Vision?

Visual object tracking and segmentation are becoming fundamental tasks for understanding human activities in egocentric vision. Recent research has benchmarked state-of-the-art methods and concluded that first person egocentric vision presents challenges compared to previously studied domains. However, these claims are based on evaluations conducted across significantly different scenarios. Many of the challenging characteristics attributed to egocentric vision are also present in third person videos of human-object activities. This raises a critical question: how much of the observed performance drop stems from the unique first person viewpoint inherent to egocentric vision versus the domain of human-object activities? To address this question, we introduce a new benchmark study designed to disentangle such factors. Our evaluation strategy enables a more precise separation of challenges related to the first person perspective from those linked to the broader domain of human-object activity understanding. By doing so, we provide deeper insights into the true sources of difficulty in egocentric tracking and segmentation, facilitating more targeted advancements on this task.

cs.CV↗

arXiv2024

Online Episodic Memory Visual Query Localization with Egocentric Streaming Object Memory

Episodic memory retrieval enables wearable cameras to recall objects or events previously observed in video. However, existing formulations assume an "offline" setting with full video access at query time, limiting their applicability in real-world scenarios with power and storage-constrained wearable devices. Towards more application-ready episodic memory systems, we introduce Online Visual Query 2D (OVQ2D), a task where models process video streams online, observing each frame only once, and retrieve object localizations using a compact memory instead of full video history. We address OVQ2D with ESOM (Egocentric Streaming Object Memory), a novel framework integrating an object discovery module, an object tracking module, and a memory module that find, track, and store spatio-temporal object information for efficient querying. Experiments on Ego4D demonstrate ESOM's superiority over other online approaches, though OVQ2D remains challenging, with top performance at only ~4% success. ESOM's accuracy increases markedly with perfect object tracking (31.91%), discovery (40.55%), or both (81.92%), underscoring the need of applied research on these components.

cs.CV↗

arXiv2024

LowFormer: Hardware Efficient Design for Convolutional Transformer Backbones

Research in efficient vision backbones is evolving into models that are a mixture of convolutions and transformer blocks. A smart combination of both, architecture-wise and component-wise is mandatory to excel in the speedaccuracy trade-off. Most publications focus on maximizing accuracy and utilize MACs (multiply accumulate operations) as an efficiency metric. The latter however often do not measure accurately how fast a model actually is due to factors like memory access cost and degree of parallelism. We analyzed common modules and architectural design choices for backbones not in terms of MACs, but rather in actual throughput and latency, as the combination of the latter two is a better representation of the efficiency of models in real applications. We applied the conclusions taken from that analysis to create a recipe for increasing hardware-efficiency in macro design. Additionally we introduce a simple slimmed-down version of MultiHead Self-Attention, that aligns with our analysis. We combine both macro and micro design to create a new family of hardware-efficient backbone networks called LowFormer. LowFormer achieves a remarkable speedup in terms of throughput and latency, while achieving similar or better accuracy than current state-of-the-art efficient backbones. In order to prove the generalizability of our hardware-efficient design, we evaluate our method on GPU, mobile GPU and ARM CPU. We further show that the downstream tasks object detection and semantic segmentation profit from our hardware-efficient architecture. Code and models are available at https://github.com/ altair199797/LowFormer.

cs.CV↗

arXiv2023

Tracking Skiers from the Top to the Bottom

Skiing is a popular winter sport discipline with a long history of competitive events. In this domain, computer vision has the potential to enhance the understanding of athletes' performance, but its application lags behind other sports due to limited studies and datasets. This paper makes a step forward in filling such gaps. A thorough investigation is performed on the task of skier tracking in a video capturing his/her complete performance. Obtaining continuous and accurate skier localization is preemptive for further higher-level performance analyses. To enable the study, the largest and most annotated dataset for computer vision in skiing, SkiTB, is introduced. Several visual object tracking algorithms, including both established methodologies and a newly introduced skier-optimized baseline algorithm, are tested using the dataset. The results provide valuable insights into the applicability of different tracking methods for vision-based skiing analysis. SkiTB, code, and results are available at https://machinelearning.uniud.it/datasets/skitb.

cs.CV↗

arXiv2023

Visualizing Skiers' Trajectories in Monocular Videos

Trajectories are fundamental to winning in alpine skiing. Tools enabling the analysis of such curves can enhance the training activity and enrich broadcasting content. In this paper, we propose SkiTraVis, an algorithm to visualize the sequence of points traversed by a skier during its performance. SkiTraVis works on monocular videos and constitutes a pipeline of a visual tracker to model the skier's motion and of a frame correspondence module to estimate the camera's motion. The separation of the two motions enables the visualization of the trajectory according to the moving camera's perspective. We performed experiments on videos of real-world professional competitions to quantify the visualization error, the computational efficiency, as well as the applicability. Overall, the results achieved demonstrate the potential of our solution for broadcasting media enhancement and coach assistance.

cs.CV↗

arXiv2022

Visual Object Tracking in First Person Vision

The understanding of human-object interactions is fundamental in First Person Vision (FPV). Visual tracking algorithms which follow the objects manipulated by the camera wearer can provide useful information to effectively model such interactions. In the last years, the computer vision community has significantly improved the performance of tracking algorithms for a large variety of target objects and scenarios. Despite a few previous attempts to exploit trackers in the FPV domain, a methodical analysis of the performance of state-of-the-art trackers is still missing. This research gap raises the question of whether current solutions can be used ``off-the-shelf'' or more domain-specific investigations should be carried out. This paper aims to provide answers to such questions. We present the first systematic investigation of single object tracking in FPV. Our study extensively analyses the performance of 42 algorithms including generic object trackers and baseline FPV-specific trackers. The analysis is carried out by focusing on different aspects of the FPV setting, introducing new performance measures, and in relation to FPV-specific tasks. The study is made possible through the introduction of TREK-150, a novel benchmark dataset composed of 150 densely annotated video sequences. Our results show that object tracking in FPV poses new challenges to current visual trackers. We highlight the factors causing such behavior and point out possible research directions. Despite their difficulties, we prove that trackers bring benefits to FPV downstream tasks requiring short-term object tracking. We expect that generic object tracking will gain popularity in FPV as new and FPV-specific methodologies are investigated.

cs.CV↗

arXiv2022

CoCoLoT: Combining Complementary Trackers in Long-Term Visual Tracking

How to combine the complementary capabilities of an ensemble of different algorithms has been of central interest in visual object tracking. A significant progress on such a problem has been achieved, but considering short-term tracking scenarios. Instead, long-term tracking settings have been substantially ignored by the solutions. In this paper, we explicitly consider long-term tracking scenarios and provide a framework, named CoCoLoT, that combines the characteristics of complementary visual trackers to achieve enhanced long-term tracking performance. CoCoLoT perceives whether the trackers are following the target object through an online learned deep verification model, and accordingly activates a decision policy which selects the best performing tracker as well as it corrects the performance of the failing one. The proposed methodology is evaluated extensively and the comparison with several other solutions reveals that it competes favourably with the state-of-the-art on the most popular long-term visual tracking benchmarks.

cs.CV↗

arXiv2021

Video-Based Reconstruction of the Trajectories Performed by Skiers

Trajectories are fundamental in different skiing disciplines. Tools enabling the analysis of such curves can enhance the training activity and enrich the broadcasting contents. However, the solutions currently available are based on geo-localized sensors and surface models. In this short paper, we propose a video-based approach to reconstruct the sequence of points traversed by an athlete during its performance. Our prototype is constituted by a pipeline of deep learning-based algorithms to reconstruct the athlete's motion and to visualize it according to the camera perspective. This is achieved for different skiing disciplines in the wild without any camera calibration. We tested our solution on broadcast and smartphone-captured videos of alpine skiing and ski jumping professional competitions. The qualitative results achieved show the potential of our solution.

cs.CV↗

arXiv2021

Is First Person Vision Challenging for Object Tracking?

Understanding human-object interactions is fundamental in First Person Vision (FPV). Tracking algorithms which follow the objects manipulated by the camera wearer can provide useful cues to effectively model such interactions. Despite a few previous attempts to exploit trackers in FPV applications, a methodical analysis of the performance of state-of-the-art visual trackers in this domain is still missing. In this short paper, we provide a recap of the first systematic study of object tracking in FPV. Our work extensively analyses the performance of recent and baseline FPV trackers with respect to different aspects. This is achieved through TREK-150, a novel benchmark dataset composed of 150 densely annotated video sequences. The results suggest that more research efforts should be devoted to this problem so that tracking could benefit FPV tasks. The full version of this paper is available at arXiv:2108.13665.

cs.CV↗

arXiv2021

Is First Person Vision Challenging for Object Tracking?

Understanding human-object interactions is fundamental in First Person Vision (FPV). Tracking algorithms which follow the objects manipulated by the camera wearer can provide useful cues to effectively model such interactions. Visual tracking solutions available in the computer vision literature have significantly improved their performance in the last years for a large variety of target objects and tracking scenarios. However, despite a few previous attempts to exploit trackers in FPV applications, a methodical analysis of the performance of state-of-the-art trackers in this domain is still missing. In this paper, we fill the gap by presenting the first systematic study of object tracking in FPV. Our study extensively analyses the performance of recent visual trackers and baseline FPV trackers with respect to different aspects and considering a new performance measure. This is achieved through TREK-150, a novel benchmark dataset composed of 150 densely annotated video sequences. Our results show that object tracking in FPV is challenging, which suggests that more research efforts should be devoted to this problem so that tracking could benefit FPV tasks.

cs.CV↗

arXiv2021

Weakly-Supervised Domain Adaptation of Deep Regression Trackers via Reinforced Knowledge Distillation

Deep regression trackers are among the fastest tracking algorithms available, and therefore suitable for real-time robotic applications. However, their accuracy is inadequate in many domains due to distribution shift and overfitting. In this paper we overcome such limitations by presenting the first methodology for domain adaption of such a class of trackers. To reduce the labeling effort we propose a weakly-supervised adaptation strategy, in which reinforcement learning is used to express weak supervision as a scalar application-dependent and temporally-delayed feedback. At the same time, knowledge distillation is employed to guarantee learning stability and to compress and transfer knowledge from more powerful but slower trackers. Extensive experiments on five different robotic vision domains demonstrate the relevance of our methodology. Real-time speed is achieved on embedded devices and on machines without GPUs, while accuracy reaches significant results.

cs.CV↗

arXiv2020

Tracking-by-Trackers with a Distilled and Reinforced Model

Visual object tracking was generally tackled by reasoning independently on fast processing algorithms, accurate online adaptation methods, and fusion of trackers. In this paper, we unify such goals by proposing a novel tracking methodology that takes advantage of other visual trackers, offline and online. A compact student model is trained via the marriage of knowledge distillation and reinforcement learning. The first allows to transfer and compress tracking knowledge of other trackers. The second enables the learning of evaluation measures which are then exploited online. After learning, the student can be ultimately used to build (i) a very fast single-shot tracker, (ii) a tracker with a simple and effective online adaptation mechanism, (iii) a tracker that performs fusion of other trackers. Extensive validation shows that the proposed algorithms compete with real-time state-of-the-art trackers.

cs.CV↗