Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks308 from public data
- 876
- 597
- Learning Deep Representations of Appearance and Motion for Anomalous Event Detection550
This work proposes Appearance and Motion DeepNet (AMDN) which utilizes deep neural networks to automatically learn feature representations, and introduces a novel double fusion framework, combining both the benefits of traditional early fusion and late fusion strategies.
- Detecting anomalous events in videos by learning deep representations of appearance and motion474
A novel double fusion framework is introduced, combining the benefits of traditional early fusion and late fusion strategies, which is extensively evaluated on publicly available video surveillance datasets including UCSD pedestian, Subway, and Train, showing competitive performance with respect to state of the art approaches.
- Multi-scale Continuous CRFs as Sequential Deep Networks for Monocular Depth Estimation466
This paper addresses the problem of depth estimation from a single still image by designing a novel CNN implementation of mean-field updates for continuous CRFs and demonstrates the effectiveness of the proposed approach and establishes new state of the art results on publicly available datasets.
- Animating Arbitrary Objects via Deep Motion Transfer400
This paper introduces a novel deep learning framework for image animation that generates a video in which the target object is animated according to the driving sequence through a deep architecture that decouples appearance and motion information.
- Modeling the Background for Incremental Learning in Semantic Segmentation386
This work revisits classical incremental learning methods, and proposes a new distillation-based framework which explicitly accounts for a semantic distribution shift, and introduces a novel strategy to initialize classifier's parameters, thus preventing biased predictions toward the background class.
- Structured Attention Guided Convolutional Neural Fields for Monocular Depth Estimation348
This paper introduces a novel approach for monocular depth estimation which is competitive with previous methods on the KITTI benchmark and outperforms the state of the art on the NYU Depth V2 dataset.
- 340
- AutoDIAL: Automatic Domain Alignment Layers337
Opposite to previous works which define a priori in which layers adaptation should be performed, this method is able to automatically learn the degree of feature alignment required at different levels of the deep network.
- A Unified Objective for Novel Class Discovery281
A UNified Objective function (UNO) for discovering novel classes, with the explicit purpose of favoring synergy between supervised and unsupervised learning, and outperforms the state of the art on several benchmarks.
- Transformer-Based Attention Networks for Continuous Pixel-Wise Prediction260
This paper proposes TransDepth, an architecture that benefits from both convolutional neural networks and transformers that applies transformers to pixel-wise prediction problems involving continuous labels and achieves state-of-theart performance on three challenging datasets.
- Self-Supervised Models are Continual Learners246
It is shown that self-supervised loss functions can be seamlessly converted into distillation mechanisms for CL by adding a predictor network that maps the current state of the repre-sentations to their past state.
- Learning Cross-Modal Deep Representations for Robust Pedestrian Detection237
The proposed approach outperforms the state-of-the-art on the challenging KAIST multispectral pedestrian dataset and it is competitive with previous methods on the popular Caltech dataset.
- Neighborhood Contrastive Learning for Novel Class Discovery211
This paper addresses Novel Class Discovery (NCD), the task of unveiling new classes in a set of unlabeled samples given a labeled dataset with known classes, and builds a new framework, named Neighborhood Contrastive Learning (NCL), to learn discriminative representations that are important to clustering performance.
- Unsupervised Adversarial Depth Estimation Using Cycled Generative Networks193
A novel unsupervised deep learning approach for predicting depth maps and showing that the depth estimation task can be effectively tackled within an adversarial learning framework is presented.
- Unsupervised Domain Adaptation Using Feature-Whitening and Consensus Loss182
This work proposes domain alignment layers which implement feature whitening for the purpose of matching source and target feature distributions, and leverage the unlabeled target data by proposing the Min-Entropy Consensus loss, which regularizes training while avoiding the adoption of many user-defined hyper-parameters.
- Precise widening operators for convex polyhedra181
A framework for the systematic definition of new widening operators that are never less precise than a given widening is presented and an improvement to the well-known widening delay technique is suggested that allows one to gain precision while preserving its overall simplicity.
- 3D Object Detection From Images for Autonomous Driving: A Survey180
This paper provides the first comprehensive survey of this novel and continuously growing research field, summarizing the most commonly used pipelines for image-based 3D detection and deeply analyzing each of their components and proposing two new taxonomies to organize the state-of-the-art methods into different categories.
- Best Sources Forward: Domain Generalization through Source-Specific Nets176
This work designs a deep network with multiple domain-specific classifiers, each associated to a source domain, and introduced a domain agnostic component supporting the final classifier.
- SALSA: A Novel Dataset for Multimodal Group Behavior Analysis170
This work proposes SALSA, a novel dataset facilitating multimodal and Synergetic sociAL Scene Analysis, and shows how the recorded multiple cues synergetically aid automatic analysis of social interactions.
- Predicting and Understanding Urban Perception with Convolutional Neural Networks167
A novel approach for predicting the perceived safety of a scene from Google Street View Images using a Convolutional Neural Network (CNN), significantly improving the accuracy of predictions over previous methods.
- 165
- Boosting Domain Adaptation by Discovering Latent Domains161
A novel Convolutional Neural Network (CNN) architecture is introduced which automatically discovers latent domains in visual datasets and exploits this information to learn robust target classifiers and outperforms state-of-the-art multi-source DA methods by a large margin.
- 161
- Possibly Not Closed Convex Polyhedra and the Parma Polyhedra Library159
The Parma Polyhedra Library is presented, a new, robust and complete implementation of NNC convex polyhedra, concentrating on the distinctive features of the library and on the novel theoretical underpinnings.
- 155
- Refine and Distill: Exploiting Cycle-Inconsistency and Knowledge Distillation for Unsupervised Monocular Depth Estimation149
A novel self-supervised deep model for estimating depth maps which outperforms state of the art unsupervised methods on the KITTI benchmark and exploits the inconsistency between the original and the reconstructed input frame.
- We are not All Equal144
This work proposes a new transfer learning method to compute personalized models without labeled target data based on learning multiple person-specific classifiers for a set of source subjects and then directly transfer knowledge about the parameters of these classifiers to the target individual.
- Harnessing Large Language Models for Training-Free Video Anomaly Detection139
This paper proposes LAnguage-based VAD (LAVAD), a method tackling VAD in a novel, training-free paradigm, exploiting the capabilities of pre-trained large language models (LLMs) and existing vision-language models (VLMs).
- Learning Deep Structured Multi-Scale Features using Attention-Gated CRFs for Contour Prediction134
A hierarchical deep model is introduced which produces more rich and complementary representations and the novel Attention-Gated Conditional Random Fields (AG-CRFs) are proposed to refine and robustly fuse the representations learned at different scales.
- 132
- 128
- Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis122
Snap Video is built, a video-first model that systematically addresses the challenges of motion fidelity, visual quality and scalability in the video generation domain and proposes a new transformer-based architecture that trains 3.31 times faster than U-Nets (and is ∼4.5 faster at inference).
- Towards Recognizing Unseen Categories in Unseen Domains121
CuMix (CurriculumMixup for recognizing unseen categories in unseen domains), a holistic algorithm to tackle ZSL, DG and ZSL+DG, to simulate the test-time domain and semantic shift using images and features from unseen domains and categories generated by mixing up the multiple source domains and Categories available during training.
- Space speaks: towards socially and personality aware visual surveillance121
A camera based monitoring system that explores the relationship between proxemics, visual attention and personality traits during interaction and reports on experimental results in a natural scenario where people engage in a party.
- 120
- Monocular Depth Estimation Using Multi-Scale Continuous CRFs as Sequential Deep Networks112
This paper addresses the problem of monocular depth estimation from a single still image by designing a novel CNN implementation of mean-field updates for continuous CRFs and proposes two different variations, one based on a cascade of multiple CRFs, the other on a unified graphical model.
- Automated defect detection in uniform and structured fabrics using Gabor filters and PCA107
The results show that the algorithm outperforms previous approaches in most cases, achieving a detection rate of 98.8% and a false alarm rate as low as 0.20-0.37%, whereas for heavily structured yarns misdetection rate can be as high as 5%.
- Uncertainty-Aware Contrastive Distillation for Incremental Semantic Segmentation106
A novel distillation framework, Uncertainty-aware Contrastive Distillation, that takes into account all the images in a mini-batch, enforcing similarity between features associated to all the pixels from the same classes, and pulling apart those corresponding to pixels from different classes is proposed.
- 97
- 96
- 94
- 90
- 89
- Playable Video Generation85
A novel framework for PVG is proposed that is trained in a self-supervised manner on a large dataset of unlabelled videos where the predicted action labels act as bottleneck and the network is constrained to learn a rich action space using a reconstruction loss on the generated video.
- Analyzing Free-standing Conversational Groups80
This paper introduces a framework able to fuse multimodal data emanating from a combination of distributed and wearable sensors, taking into account the temporal consistency, the head/body coupling and the noise inherent to the scenario.
- A smart watch-based gesture recognition system for assisting people with visual impairments80
A novel approach for gesture recognition which is based on global alignment kernels is shown to be effective in the challenging scenario of user independent recognition and is embedded into a system targeted to visually impaired which will also integrate several other modules.
- ConfMix: Unsupervised Domain Adaptation for Object Detection via Confidence-based Mixing78
ConfMix is the first method that introduces a sample mixing strategy based on region-level detection confidence for adaptive object detector learning, and proposes a novel pseudo labelling scheme that progressively filters the pseudo target detections using the confidence metric that varies from a loose to strict manner along the training.
- Curriculum Graph Co-Teaching for Multi-Target Domain Adaptation77
This paper develops a co-teaching strategy with the dual classifier head that is assisted by curriculum learning to obtain more reliable pseudo-labels and proposes Domain-aware Curriculum Learning (DCL), a sequential adaptation strategy that first adapts on the easier target domains, followed by the harder ones.
- Towards Generalization Across Depth for Monocular 3D Object Detection77
This work introduces MoVi-3D, a novel, single-stage deep architecture for monocular 3D object detection which builds upon a novel approach which leverages geometrical information to generate virtual views where the object appearance is normalized with respect to distance.
- AdaGraph: Unifying Predictive and Continuous Domain Adaptation Through Graphs77
This paper presents the first deep architecture that tackles predictive domain adaptation, able to leverage over the information brought by the auxiliary domains through a graph, and presents a simple yet effective strategy that allows to take advantage of the incoming target data at test time, in a continuous domain adaptation scenario.
- Online Continual Learning Under Extreme Memory Constraints76
Batch-level Distillation (BLD) is proposed, a regularization-based CL approach, which effectively balances stability and plasticity in order to learn from data streams, while preserving the ability to solve old tasks through distillation.
- Every Smile is Unique: Landmark-Guided Diverse Smile Generation71
This work proposes a novel deep learning architecture named Conditional Multi-Mode Network (CMM-Net), which explicitly exploits facial landmarks for generating smile sequences and demonstrates the effectiveness of this architecture in generating realistic videos of multiple smile expressions.
- 67
- 66
- 63
- Kitting in the Wild through Online Domain Adaptation63
This work presents a new visual dataset for the kitting task, and benchmark the performance of the algorithm on the proposed dataset, showing its capability to fill the gap between the performances of a standard architecture and its counterpart adapted offline to the given target domain.
- StyLIP: Multi-Scale Style-Conditioned Prompt Learning for CLIP-based Domain Generalization62
This study proposes StyLIP, a novel approach for Domain Generalization (DG) that enhances CLIP’s classification performance across domains by introducing a set of style projectors that directly learn the domain-specific prompt tokens from the extracted multi-scale style features.
- Exploring Transfer Learning Approaches for Head Pose Classification from Multi-view Surveillance Images62
This paper examines the use of transfer learning for efficient multi-view head pose classification with minimal target training data under three challenging situations: (i) where the range of head poses in the source and target images is different, (ii) where source images captured a stationary person while target images capture a moving person whose facial appearance varies under motion due to changing perspective, scale.
- 61
- 60
- Robust Place Categorization With Deep Domain Generalization60
This paper proposes a novel deep learning framework for domain generalization that exploits recent advances in deep domain adaptation and design a convolutional neural network architecture with novel layers performing a weighted version of batch normalization.
- Exploring Diffusion Models for Unsupervised Video Anomaly Detection57
This paper investigates the performance of diffusion models for video anomaly detection (VAD) within the most challenging but also the most operational scenario in which the data annotations are not used, using a diffusion model and examining its parameters’ influence to present guidance for VAD in surveillance scenarios.
- Semi-supervised learning made simple with self-supervised clustering56
This paper introduces a multi-task framework merging a supervised objective using ground-truth labels and a self-supervised objective relying on clustering assignments with a single cross-entropy loss and provides empirical evidence that this approach is highly effective and achieves state-of-the-art performance on CI-FAR100 and ImageNet.
- Adding New Tasks to a Single Network with Weight Transformations Using Binary Masks56
This work shows that with the generalization of this approach it is possible to achieve significantly higher levels of adaptation to new tasks, enabling the approach to compete with fine tuning strategies by requiring slightly more than 1 bit per network parameter per additional task.
- 55
- 54
- Unsupervised Domain Adaptation for Personalized Facial Emotion Recognition53
A personalization approach in which only unlabeled target-specific data are required and a simplified regression framework is presented which achieves the same or even slightly superior experimental results but it is much easier to reproduce.
- SF-UDA 3D : Source-Free Unsupervised Domain Adaptation for LiDAR-Based 3D Object Detection51
SF-UDA3D outperforms both previous domain adaptation techniques based on features alignment and state-of-the-art 3D object detection methods which additionally use few-shot target annotations or target annotation statistics.
- Just DIAL: DomaIn Alignment Layers for Unsupervised Domain Adaptation51
This work acts at the distribution level through the introduction of \emph{DomaIn Alignment Layers} (\DIAL), able to match the observed source and target data distributions to a reference one.
- Evaluation of non-geometric methods for visual odometry50
The experiments show that non-geometric VO is a good alternative, or addition, to standard VO systems, and stress the advantages of non- geometrical VO as an alternative or an addition to standard geometric methods.
- 50
- Object-aware Gaze Target Detection49
A Transformer-based architecture that automatically detects objects in the scene to build associations between every head and the gazed-head/object, resulting in a comprehensive, explainable gaze analysis.
- 49
- Budget-Aware Adapters for Multi-Domain Learning46
Budget-Aware Adapters are introduced that select the most relevant feature channels to better handle data from a novel domain and propose a novel strategy to automatically adjust the computational complexity of the network.
- Overlap-guided Gaussian Mixture Models for Point Cloud Registration45
A novel overlap-guided probabilistic registration approach that computes the optimal transformation from matched Gaussian Mixture Model parameters, and achieves superior registration accuracy and efficiency when handling point clouds with partial overlap and different densities on synthetic and real-world datasets.
- 45
- 45
- 45
- 45
- Continual Attentive Fusion for Incremental Learning in Semantic Segmentation44
A novel attentive feature distillation approach to mitigate catastrophic forgetting while accounting for semantic spatial- and channel-level dependencies is introduced and a continual attentive fusion structure is proposed, which takes advantage of the attention learned from the new and the old tasks while learning features for the new task.
- 43
- 42
- 41
- 40
- Motion-supervised Co-Part Segmentation39
This work proposes a self-supervised deep learning method for co-part segmentation that develops the idea that motion information inferred from videos can be leveraged to discover meaningful object parts.
- Are we Missing Confidence in Pseudo-LiDAR Methods for Monocular 3D Object Detection?38
3D confidence estimation techniques derived from RGB-only 3D detection approaches can be successfully integrated into the framework and improved performance can be obtained with a newly designed 3D confidence measure, leading to state-of-the-art performance on the KITTI3D benchmark.
- 37
- Compositional Semantic Mix for Domain Adaptation in Point Cloud Segmentation36
This study introduces compositional semantic mixing for point cloud domain adaptation, representing the first unsupervised domain adaptation technique for pointCloud segmentation based on semantic and geometric sample mixing and presents a two-branch symmetric network architecture capable of concurrently processing point clouds from a source domain and a target domain.
- Playable Environments: Video Manipulation in Space and Time36
This work presents Playable Environments-a new representation for interactive video generation and manipulation in space and time, which allows the user to move objects in 3D while generating a video by providing a sequence of desired actions.
- Novel Class Discovery for 3D Point Cloud Semantic Segmentation35
A new method for NCD based on online clustering that exploits uncertainty quantification to produce prototypes for pseudo-labelling the points of the novel classes is presented, showing that it can significantly outperform the baseline.
- Joint Estimation of Human Pose and Conversational Groups from Social Scenes34
A novel framework for jointly estimating head, body orientations of targets and conversational groups called F-formations from social scenes is presented, and a weakly-supervised learning algorithm for joint inference is proposed.
- How to Make an Image More Memorable?33
This work shows that it is possible to automatically retrieve the best style seeds for a given image, thus remarkably reducing the number of human attempts needed to find a good match.
- Learning Deep NBNN Representations for Robust Place Categorization33
Experimental results show that the proposed algorithm outperforms previous methods based on pretrained CNN models and that, when employed in challenging robot place recognition tasks, it is robust to occlusions, environmental and sensor changes.
- Learning Depth-Aware Deep Representations for Robotic Perception33
This paper shows that the performance of deep architectures can be boosted by introducing DaConv, a novel, general-purpose CNN block which exploits depth to learn scale-aware feature representations in RGB-D images.
- Progressive Fusion for Unsupervised Binocular Depth Estimation Using Cycled Networks32
This paper introduces a new network architecture, named Progressive Fusion Network (PFN), that is specifically designed for binocular stereo depth estimation, based on a multi-scale refinement strategy that combines the information provided by both stereo views.
- Inferring Latent Domains for Unsupervised Deep Domain Adaptation32
A novel deep architecture which addresses the problem of UDA by automatically discovering latent domains in visual datasets and exploiting this information to learn robust target classifiers is introduced, showing that it outperforms state-of-the-art domain adaptation methods.
- 32
- Rethinking Few-Shot Adaptation of Vision-Language Models in Two Stages31
This work starts by analyzing the learning dynamics of PEFT techniques when trained on few-shot data from only a subset of categories, and shows that such dynamics naturally splits into two distinct phases: task-level feature extraction and specialization to the available concepts.
- Test-Time Zero-Shot Temporal Action Localization29
A novel method that performs Test-Time adaptation for Temporal Action Localization (T3AL), which adapts a pre-trained Vision and Language Model (VLM) and demonstrates that T3AL significantly outperforms zero-shot baselines based on state-of-the-art VLMs, confirming the benefit of a test-time adaptation approach.
- Learning Contours for Automatic Annotations of Mountains Pictures on a Smartphone29
An AR application for mobile phones to augment pictures of mountainous landscapes with geo-referenced data (e.g. the peaks' names, positions of mountain dews or hiking tracks) is presented, based on a novel approach for image-to-world registration, which exploits different information collected with on-board sensors.
- 29
- A novel machine learning-based framework for the water quality parameters prediction using hybrid long short-term memory and locally weighted scatterplot smoothing methods28
This work proposes a novel hybrid approach for water quality parameters prediction combining DL with data smoothing method, which demonstrates that LSTM outperforms the baseline method, with an R2 score up to 0.9998 and RMSE of 0.0230 on the test set for dissolved oxygen.
- Promptable Game Models: Text-guided Game Simulation via Masked Diffusion Models28
A Promptable Game Model (PGM), which significantly outperforms existing neural video game simulators in terms of rendering quality and unlocks applications beyond the capabilities of the current state-of-the-art.
- Probabilistic Graph Attention Network with Conditional Kernels for Pixel-Wise Prediction28
Quantitative and qualitative results demonstrate the effectiveness of the proposed latent AG-CRF model and the overall probabilistic graph attention network with feature conditional kernels for structured feature learning and pixel-wise prediction.
- Inferring painting style with multi-task dictionary learning28
This paper presents a novel dictionary learning approach to automatically uncover the artistic style from paintings, by automatically decoupling style-specific and artist-specific patterns, and shows that this approach significantly outperforms state-of-the-art methods.
- Democratizing Fine-grained Visual Recognition with Large Language Models27
This work proposes Fine-grained Semantic Category Reasoning (FineR) that internally leverages the world knowledge of large language models (LLMs) as a proxy in order to reason about fine-grained category names.
- A soft nearest-neighbor framework for continual semi-supervised learning27
This paper proposes an approach for continual semi-supervised learning—a setting where not all the data samples are labeled, and leverages the power of nearest-neighbor classifiers to nonlinearly partition the feature space and flexibly model the underlying data distribution thanks to its non-parametric nature.
- AutoLabel: CLIP-based framework for Open-Set Video Domain Adaptation27
This work proposes AutoLabel that automatically discovers and generates object-centric compositional candidate target-private class names and shows that CLIP when equipped with AutoLabel can satisfactorily reject the target- private instances, thereby facilitating better alignment between the shared classes of the two domains.
- 27
- Multimodal Emotion Recognition with Modality-Pairwise Unsupervised Contrastive Loss27
This work considers discrete emotions, and as modalities text, audio and vision are used, and end-to-end feature learning approach is the first attempt in MER literature based on contrastive loss between pairwise modalities.
- Walking Your LiDOG: A Journey Through Multiple Domains for LiDAR Semantic Segmentation26
This work designs the first experimental setup for studying domain generalization (DG) for LiDAR semantic segmentation (DG-LSS), and proposes the first method specifically designed for DG-LSS, which obtains 34.88 mIoU on the target domain, outperforming all baselines.
- Learning How to Smile: Expression Video Generation With Conditional Adversarial Recurrent Nets26
A novel deep generative model able to produce face videos from a given image of a neutral face and a label indicating a specific facial expression, e.g. spontaneous smile is proposed.
- Knowledge is Never Enough: Towards Web Aided Deep Open World Recognition26
The proposed deep network, based on a deep extension of a non-parametric model, detects whether a perceived object belongs to the set of categories known by the system and learns it without the need to retrain the whole system from scratch.
- 26
- 26
- Boosting Deep Open World Recognition by Clustering25
This work shows how it can boost the performance of deep open world recognition algorithms by means of a new loss formulation enforcing a global to local clustering of class-specific features, and proposes a strategy to learn class- specific rejection thresholds.
- Online Multitask Learning for Machine Translation Quality Estimation25
The effectiveness of the approach in a regression task (HTER prediction), in which online multitask learning outperforms the competitive online single-task and pooling methods used for comparison, indicates the feasibility of integrating in a CAT tool a single QE component capable to simultaneously serve multiple translation jobs involving different domains and users.
- 24
- Multimodal behavior analysis in the wild: An introduction24
This book describes some recent research efforts in the area of human behavior analysis, presenting methodologies for extracting behavioral cues at different levels, with special emphasis given to recent approaches considering multimodal data to robustly extract behavioral information in real-world settings.
- SHiNe: Semantic Hierarchy Nexus for Open-Vocabulary Object Detection23
Semantic Hierarchy Nexus (SHiNe), a novel classifier that uses semantic knowledge from class hierarchies that enhances robustness across diverse vocabulary granularities and can be seamlessly integrated with any off-the-shelf OvOD detector, without incurring additional computational overhead during inference.
- The Unreasonable Effectiveness of Large Language-Vision Models for Source-free Video Domain Adaptation23
This work showcases the unreasonable effectiveness of integrating LLVMs for SFVUDA by devising an intuitive and parameter-efficient method, which is name Domain Adaptation with Large Language-Vision models (DALL-V), that distills the world prior and complementary source model information into a student network tailored for the target.
- Cross-Paced Representation Learning With Partial Curricula for Sketch-Based Image Retrieval23
The cross-paced partial curriculum learning (CPPCL) framework is introduced, specifically designed to assess the learning pace by jointly handling data from dual sources and modality-specific prior information provided in the form of partial curricula.
- Viraliency: Pooling Local Virality22
This work introduces a pooling layer that learns the size of the support area to be averaged: the learned top-N average (LENA) pooling, and assesses the effectiveness of the LENA layer by appending it on top of a convolutional siamese architecture and evaluates its performance on the task of predicting and localizing virality.
- 22
- 21
- 21
- Increasing Image Memorability with Neural Style Transfer21
An approach based on an editing-by-applying-filters paradigm is introduced, which proposes to automatically retrieve a set of style images that, applied to the input image through a neural style transfer algorithm, provide the highest increase in memorability.
- Quantum Multi-Model Fitting20
This paper shows that the latter case can significantly benefit from quantum hardware and proposes the first quantum approach to multimodel fitting (MMF), and formulate MMF as a problem that can be efficiently sampled by modern adiabatic quantum computers without the relaxation of the objective function.
- 20
- 20
- 20
- 19
- On the Effectiveness of LayerNorm Tuning for Continual Learning in Vision Transformers18
This work revisits and extends a simple transfer learning idea: learning task-specific normalization layers of LayerNorm, and tunes the scale and bias parameters of LayerNorm for each continual learning task, to make the classifier robust to incorrect selection of parameters during inference.
- Graph-Based Generative Face Anonymisation with Pose Preservation18
It is proved that AnonyGAN significantly outperforms the state-of-the-art methods in terms of visual naturalness, face detection and pose preservation.
- Unsupervised Domain Adaptation for Video Transformers in Action Recognition18
This approach leverages recent advances on spatio-temporal transformers to build a robust source model that better generalises to the target domain and learns domain invariant features thanks to the introduction of a novel alignment loss term derived from the Information Bottleneck principle.
- 18
- TriGAN: image-to-image translation for multi-source domain adaptation18
This paper proposes the first approach for multi-source domain adaptation (MSDA) based on generative adversarial networks, inspired by the observation that the appearance of a given image depends on three factors: the domain, the style and the content.
- An Adaptation Framework for Head-Pose Classification in Dynamic Multi-view Scenarios18
A transfer learning approach that incorporates reliability of the different face regions for pose estimation under positional variation, by transforming the target appearance to a canonical appearance corresponding to a reference scene location.
- 18
- Unsupervised Point Cloud Representation Learning by Clustering and Neural Rendering17
This work proposes an augmentation-free unsupervised approach for point clouds to learn transferable point-level features by leveraging uni-modal information for soft clustering and cross-modal information for neural rendering, and demonstrates the effectiveness of this framework in outperforming state-of-the-art techniques.
- Semantic-Guided Inpainting Network for Complex Urban Scenes Manipulation17
A novel deep learning model is proposed to alter a complex urban scene by removing a user-specified portion of the image and coherently inserting a new object (e.g. car or pedestrian) in that scene by leveraging the semantic segmentation to model the content and structure of theimage, and learn the best shape and location of the object to insert.
- Click to Move: Controlling Video Generation with Sparse Motion17
Experimental results show that C2M outperforms existing methods on two publicly available datasets, thus demonstrating the effectiveness of the GCN framework at modelling object interactions.
- Recognizing Daily Activities from First-Person Videos with Multi-task Clustering17
This paper addresses the problem of everyday activity recognition from visual data gathered from a wearable camera proposing a novel multi-task learning framework and argues that, even if label information is not provided, this approach can take advantage of the fact that the tasks of recognizing activities of daily life of multiple individuals are related.
- Dynamic Partitioned Sampling For Tracking With Discriminative Features17
A multi-cue fusion method for tracking with particle filters which relies on a novel hierarchical sampling strategy which tackles the problem of tracking in a relatively high-dimensional state space by dividing such a space into partitions, each one corresponding to a single cue, and sampling from them in a hierarchical manner.
- 16
- Self-training transformer for source-free domain adaptation16
A generic and effective framework based on Transformer, named TransDA, is proposed, for learning a generalized model for source-free domain adaptation, which can significantly improve the accuracy over the source model and can produce state-of-the-art results on all settings.
- 16
- Improved pruning strategy for radial basis function networks with dynamic decay adjustment16
A simple method is proposed, which eliminates redundant units improving generalization, and experimental results on benchmark datasets show the good performance of this approach compared to previous methods.
- MULTIFLOW: Shifting Towards Task-Agnostic Vision-Language Pruning15
This work proposes Multimodal Flow Pruning (MULTIFLOW), a first, gradient-free, pruning framework for TA-VLP where the importance of a parameter is expressed in terms of its magnitude and its information flow, by incorporating the saliency of the neurons it connects.
- A Deeply-Supervised Deconvolutional Network for Horizon Line Detection15
A novel approach for skyline detection is introduced, which adheres to a learning-based paradigm and exploits the representation power of deep architectures to improve the horizon line detection accuracy.
- Learning to Cluster Under Domain Shift14
This work addresses the problem of transferring knowledge from a source to a target domain when both source and target data have no annotations, and leverages information from data gathered from multiple source domains to build a domain-agnostic clustering model which is then refined at inference time when target data become available.
- PCA-based artifact removal algorithm for stroke detection using UWB radar imaging14
The paper shows that a PCA-based preprocessing algorithm can achieve good performance already with a computationally simple beamforming algorithm and outperform state-of-the-art preprocessing algorithms and enable a further improvement in the performance by using a multistatic approach with just a modest increase in computational complexity.
- Academic Coupled Dictionary Learning for Sketch-based Image Retrieval14
This work introduces academic learning, a paradigm in which the sample learning order is constructed both from the data, as in self-paced learning, and from partial curricula, and proposes an instantiation of this paradigm within the framework of coupled dictionary learning to address the SBIR task.
- 14
- 13
- 13
- Text-to-CAD Evaluation with CADTests12
CADTestBench is proposed, the first test-based benchmark for Text-to-CAD, based on CADTests, executable software tests that verify whether a generated CAD model satisfies the geometric and topological requirements of the input prompt.
- 12
- Diversified in-domain synthesis with efficient fine-tuning for few-shot classification12
A novel text-to-image augmentation pipeline that promotes in-domain sample diversity for better generalization in few-shot recognition, and proposes to use Low-Rank Adaptation for joint adaptation of the text and image encoders in a Vision Language Model.
- 12
- Deep Variational Generative Models for Audio-Visual Speech Separation12
The experiments show that the proposed unsupervised VAE-based method yields better separation performance than NMF-based approaches as well as a supervised deep learning-based technique.
- Performance Evaluation of Novel Microwave Imaging Algorithms for Stroke Detection using an Accurate 3D Head Model12
Comparisons with other state-of-the-art algorithms confirm the effectiveness of these novel algorithms both in terms of accuracy of stroke localization and amount of artefacts not completely removed.
- 12
- 12
- 12
- Exploiting Sparse Representations for Robust Analysis of Noisy Complex Video Scenes12
A novel unsupervised learning approach for the analysis of complex scenes which is specifically tailored to cope directly with features' noise and uncertainty, and to use short trajectory snippets to account for object motion information, in alternative to the noisy optical flow vectors used in previous works.
- Rotation Synchronization via Deep Matrix Factorization11
This paper focuses on the formulation of synchronization via neural networks, which has only recently begun to be explored in the literature, and expresses rotation synchronization in terms of matrix factorization with a deep neural network.
- 11
- 11
- 11
- Vocabulary-Free Image Classification and Semantic Segmentation10
The Vocabulary-free Image Classification (VIC) task, which aims to assign a class from an unconstrained language-induced semantic space to an input image without needing a known vocabulary, is introduced and CaSED is proposed, a training-free method that leverages a pre-trained vision-language model and an external database.
- Training-Free Personalization via Retrieval and Reasoning on Fingerprints10
This work proposes a novel method, Retrieval and Reasoning for Personalization (R2P), leveraging internal knowledge of VLMs to extract the concept fingerprint, i.e., key attributes uniquely defining the concept within its semantic class.
- Cooperative Self-Training for Multi-Target Adaptive Semantic Segmentation10
This work proposes a self-training strategy that employs pseudo-labels to induce cooperation among multiple domain-specific classifiers and employs feature stylization as an efficient way to generate image views that forms an integral part of self- training.
- Demystifying Pseudo-LiDAR for Monocular 3D Object Detection.10
The first contribution consists in analysing and showing experimentally that the validation results published by Pseudo-LiDAR-based methods are substantially biased, and the introduction of a 3D confidence prediction module is brought back up in the ranking.
- 10
- 10
- Reduced complexity RBF classifiers with support vector centres and dynamic decay adjustment10
Experimental results on several benchmark data sets show the effectiveness of the proposed RBF classifiers method, which integrates the dynamic decay adjustment algorithm with selective pruning and standard least squares techniques.
- Superpowering Open-Vocabulary Object Detectors for X-ray Vision9
RAXO is proposed, a training-free framework that repurposes off-the-shelf RGB OvOD detectors for robust X-ray detection and introduces DET-COMPASS, a new benchmark featuring bounding box annotations for over 300 object categories, enabling large-scale evaluation of OvOD in X-ray.
- One VLM to Keep it Learning: Generation and Balancing for Data-free Continual Visual Question Answering9
This work proposes the first data-free method that leverages the language generation capability of a VLM, instead of relying on external models, to produce pseudo-rehearsal data for addressing continual VQA, and generates pseudo-rehearsal data by posing previous task questions on new task data.
- Not Only Text: Exploring Compositionality of Visual Representations in Vision-Language Models9
This work demonstrates that visual embeddings of pre-trained VLMs exhibit a compositional arrangement, and evaluates the effectiveness of this property in the tasks of compositional classification and group robustness, and proposes a framework that approximates image representations with geometry-aware compositional structures in the latent space.
- Socially Pertinent Robots in Gerontological Healthcare9
This paper attempts to partially answer the question of whether or not an autonomous social interactive robot with multi-modal conversational capabilities will be useful and accepted in real-life facilities, via two waves of experiments including patients and companions in a day-care gerontological facility in Paris with a full-sized humanoid robot endowed with social and conversational interaction capabilities.
- Cluster-level pseudo-labelling for source-free cross-domain facial expression recognition9
This work proposes the first Source-Free Unsupervised Domain Adaptation (SFUDA) method for FER that exploits self-supervised pretraining to learn good feature representations from the target data and proposes a novel and robust cluster-level pseudo-labelling strategy that accounts for in-cluster statistics.
- Revisiting Viewing Graph Solvability: An Effective Approach Based on Cycle Consistency9
The main result of this paper is a method to reduce the number of unknowns by exploiting cycle consistency and an experiment on real data that shows that unsolvable graphs may appear in practice.
- Data Augmentation-free Unsupervised Learning for 3D Point Cloud Understanding9
This work proposes an augmentation-free unsupervised approach for point clouds to learn transferable point-level features via soft clustering, named SoftClu, which exploits the affiliation of points to their clusters as a proxy to enable self-training through a pseudo-label prediction task.
- Image Analysis and Processing – ICIAP 2019 20th International Conference, Trento, Italy, September 9–13, 2019, Proceedings, Part II9
This work proposes a label propagation method guided by hierarchy of partitions in the context of the marked (supervised) segmentation problem to propagate labels on a tree modelling a hierarchical representation of the image.
- Beamforming Algorithms for UWB Radar-based Stroke Detection: Trade-off Performance-Complexity9
This paper aims to provide a better understanding of this trade-off for several beamforming algorithms and radar configurations (monostatic/multistatic) applied to stroke detection.
- 9
- UWB Radar Imaging Based on Space-Time Beamforming for Stroke Detection9
The paper shows that a multistatic approach allows to detect the bleedings with a much lower number of antennas with respect to a monostatic approach.
- 9
- Position: Modular Memory is the Key to Continual Learning Agents8
This work outlines a conceptual framework for modular memory-centric architectures that leverage ICL for rapid adaptation and knowledge accumulation, and IWL for stable updates to model capabilities, charting a practical roadmap toward continually learning agents.
- UrbanVerse: Scaling Urban Simulation by Watching City-Tour Videos8
Experiments show that UrbanVerse scenes preserve real-world semantics and layouts, achieving human-evaluated realism comparable to manually crafted scenes, in urban navigation and in urban navigation policies trained in UrbanVerse exhibit scaling power laws and strong generalization.
- FedMVP: Federated Multimodal Visual Prompt Tuning for Vision-Language Models8
Extensive evaluation on 20 datasets demonstrates that FedMVP not only preserves performance on in-distribution classes and domains, but also displays higher generalizability to unseen classes and domains, surpassing state-of-the-art methods by a notable margin.
- 8
- 8
- 8
- The Robust Semantic Segmentation UNCV2023 Challenge Results8
This paper outlines the winning solutions employed in addressing the MUAD uncertainty quantification challenge held at ICCV 2023, which primarily revolved around enhancing the robustness of semantic segmentation in urban scenes under varying natural adversarial conditions.
- Interactive Neural Painting8
I-Paint is proposed, a novel method based on a conditional transformer Variational AutoEncoder (VAE) architecture with a two-stage decoder that provides good stroke suggestions and compares favorably to the state of the art.
- Detection of Berezinskii-Kosterlitz-Thouless transition via Generative Adversarial Networks8
This work trains a Generative Adversarial Network with the Entanglement Spectrum of a system bipartition, as extracted by means of Matrix Product States ansätze to identify gapless-to-gapped phase transitions in different one-dimensional models.
- Multi-frame Motion Segmentation by Combining Two-Frame Results8
This paper introduces a two-stage approach to motion segmentation on sparse and unstructured datasets involving rigid motions, motivated by multibody structure from motion, and assumes only two-frame correspondences as input without prior knowledge about trajectories.
- Boosting binary masks for multi-domain learning through affine transformations8
This work provides a general formulation of binary mask-based models for multi-domain learning by affine transformations of the original network parameters by solving the challenge of producing a single model performing a task in all the domains together.
- 8
- A Multi-task Learning Framework for Time-continuous Emotion Estimation from Crowd Annotations8
It is shown that MTL-based regression models, which simultaneously learn the relationship between low-level audio-visual features and high-level VA ratings from a collection of movie scenes, can predict VA ratings for time-contiguous snippets from each scene more effectively than scene-specific models.
- 8
- Multi-Domain Image-to-Image Translation with Adaptive Inference Graph7
This work proposes to increase the network capacity by using an adaptive graph structure that leads to an adjustable increase in number of parameters while preserving an almost constant computational cost.
- Data‐based design of robust fault detection and isolation residuals via LASSO optimization and Bayesian filtering7
A data‐based approach for the design of structured residual subsets for the robust isolation of sensor faults is proposed, and a robust recursive Bayesian Filter has been employed to process, online, the distance of the residuals from nominal fault directions, providing a fault probability for each sensor.
- 7
- 7
- Magic Moments for Structured Output Prediction7
Two novel learning algorithms are presented that rely on efficiently computing the first two moments of the scoring function over the output space, and using them to create convex objective functions for training.
- Classifier-to-Bias: Toward Unsupervised Automatic Bias Detection for Visual Classifiers6
Experiments show that C2B discovers biases beyond those of the original datasets and outperforms a recent state-of-the-art bias detection baseline that relies on task-specific annotations, being a promising first step toward addressing task-agnostic unsupervised bias detection.
- Variational Structured Attention Networks for Deep Visual Representation Learning6
VISTA-Net outperforms the state-of-the-art in multiple continuous and discrete prediction tasks, thus confirming the benefit of the proposed approach in joint structured spatial-channel attention estimation for deep representation learning.
- 6
- Image Analysis and Processing – ICIAP 20196
This work proposes a label propagation method guided by hierarchy of partitions in the context of the marked (supervised) segmentation problem to propagate labels on a tree modelling a hierarchical representation of the image.
- 6
- 6
- Jointly Estimating Interactions and Head, Body Pose of Interactors from Distant Social Scenes6
This work exploits positional and pose cues characterizing interactors and interactions to jointly infer F-formations and head, body pose of interactors in a social scene captured by surveillance cameras.
- 6
- A CNN based algorithm for retinal vessel segmentation6
A retinal vessel segmentation method based on cellular neural networks (CNNs) is proposed, characterized by a virtual template expansion obtained through a multi-step operation, which is capable to perform vessel segmentsation within short computation time.
- 5
- Ensembling Pruned Attention Heads For Uncertainty-Aware Efficient Transformers5
Hydra Ensembles is introduced, an efficient transformer-based ensemble that prunes attention heads to create diverse members and merges them via a new multi-head attention with grouped fully-connected layers to yield a compact model with inference speed close to a single network.
- Novel Class Discovery Meets Foundation Models for 3D Semantic Segmentation5
This paper introduces the novel task of NCD for point cloud semantic segmentation and presents a new NCD approach based on online clustering, uncertainty estimation, and semantic distillation and proposes a novel evaluation protocol to rigorously assess the performance of NCD in point cloud semantic segmentation.
- 5
- 4
- 4
- 4
- Coping with Pandemics4
The panel provides a valuable opportunity for people from a variety of backgrounds to exchange views on opportunities and challenges for AI multimedia in the current and post pandemics era.
- Single-Stage Monocular 3D Object Detection with Virtual Cameras4
MoVi-3D is introduced, a novel, single-stage deep architecture for monocular 3D object detection that leverages geometrical information to generate synthetic views from virtual cameras at both, training and test time, resulting in normalized object appearance with respect to distance.
- Enhancing Perceptual Attributes with Bayesian Style Generation4
This paper introduces a novel deep learning framework for synthesizing images in order to enhance a predefined perceptual attribute and demonstrates its effectiveness in two use cases, i.e. increasing image memorability and generating scary pictures.
- SALSA: A Multimodal Dataset for the Automated Analysis of Free-Standing Social Interactions4
SALSA is proposed, a dataset facilitating multimodal and Synergetic sociAL Scene Analysis, and it is shown how the recorded multiple cues synergetically aid automatic analysis of social interactions.
- Large Margin Methods for Structured Output Prediction4
Many real-life data problems require effective classification algorithms able to model structural dependencies between multiple labels and to perform classification in a multivariate setting, i.e. such that complex, non-scalar predictions must be produced in correspondence to input vectors.
- Learning to Align: A Statistical Approach4
A new machine learning approach to the inverse parametric sequence alignment problem is presented, given as training examples a set of correct pairwise global alignments, and it is shown that the Z-score is a function of the parameters and can be computed with efficient dynamic programs similar to the Needleman-Wunsch algorithm.
- Discriminative Sequence Labeling by Z-Score Optimization4
It is shown that the Z-score is a convex function of the parameters and it can be efficiently computed with dynamic programming methods, and the maximization step turns out to be solvable by a simple linear system of equations.
- Group-robust Machine Unlearning3
This work addresses the overlooked problem of non-uniformly distributed forget sets, which it refers to as group-robust machine unlearning, and presents MIU (Mutual Information-aware Machine Unlearning), the first approach for group robustness in approximate machine unlearning.
- Retrieval-enriched zero-shot image classification in low-resource domains3
This method, named CoRE (Combination of Retrieval Enrichment), enriches the representation of both query images and class prototypes by retrieving relevant textual information from large web-crawled databases, which significantly boosts classification performance by incorporating the broader contextual information relevant to the specific class.
- 3
- Shape Consistent 2D Keypoint Estimation under Domain Shift3
This paper presents a novel deep adaptation framework for estimating keypoints under domain shift, i.e, when the training and test (target) images significantly differ in terms of visual appearance.
- The RGB-D Triathlon: Towards Agile Visual Toolboxes for Robots3
This work studies the problem of learning flexible models which can handle multiple tasks in a lightweight manner in the robot vision context, proposing a new benchmark, the RGB-D Triathlon, and evaluating state of the art algorithms in this novel challenging scenario.
- 3
- Exploring Multitask and Transfer Learning Algorithms for Head Pose Estimation in Dynamic Multiview Scenarios3
This chapter reviews recent works addressing HPE under target motion and examines the use of transfer learning and multitask learning for HPE, which can address facial appearance variations under motion.
- An automatic image-to-DEM alignment approach for annotating mountains pictures on a smartphone3
This paper presents a novel approach for image-to-world registration which jointly relies on inertial and visual sensors, and demonstrates that the proposed registration method guarantees an increased accuracy with respect to competing methods and it is computationally efficient when implemented on a smartphone.
- Transferring knowledge across robots: A risk sensitive approach3
This paper considers the scenario where a robot equipped with a monocular camera explores a new environment and introduces a novel transfer learning approach, which casts the visual place recognition problem within a transfer learning framework.
- Sorting Atomic Activities for Discovering Spatio-temporal Patterns in Dynamic Scenes3
A novel non-object centric approach for discovering activity patterns in dynamic scenes by dividing the video into clips, compute clip histograms and cluster them to discover spatio-temporal patterns using a recently proposed clustering algorithm.
- 3
- DataComp-VLM: Improved Open Datasets for Vision-Language Models2
It is found 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.
- Thinking Past the Answer: Evaluating Harmful Overthinking in Large Reasoning Models2
A prefix-level trajectory evaluation protocol grounded in reasoning sufficiency is introduced, defining the minimum reasoning budget required for a model to first generate the correct answer, and it is found that many instances considered reasoning-intensive require surprisingly little reasoning.
- MAMBO: High-Resolution Generative Approach for Mammography Images2
Experiments assess MAMBO's capabilities in image generation, super-resolution, and anomaly segmentation, highlighting its potential to enhance mammography analysis for more accurate diagnoses and earlier lesion detection.
- Dynamic Scoring with Enhanced Semantics for Training-Free Human-Object Interaction Detection2
A novel training-free HOI detection framework for Dynamic Scoring with enhanced semantics (dysco) that effectively utilizes textual and visual interaction representations within a multimodal registry, enabling robust and nuanced interaction understanding.
- 2
- Exploring Fine-Grained Retail Product Discrimination with Zero-Shot Object Classification Using Vision-Language Models2
A novel ensemble approach that integrates embeddings from CLIP and DINOv2 with dimensionality reduction techniques to enhance classification performance and introduces a class adaptation method that utilizes visual prototyping with limited samples in scenarios with scarce labeled data, addressing a critical need in retail environments where product variety frequently changes.
- 2
- 2
- 2
- 2
- Emerging Topics in Learning from Noisy and Missing Data2
This tutorial will present several recent methodologies addressing different visual tasks under the assumption of noisy, weakly annotated data sets, including methods able to deal with noisy,weak or partial annotations.
- 1
- 1
- Can Text-to-Video Generation help Video-Language Alignment?1
Experiments show that, on average, SynViTa improves over existing methods on VideoCon test sets and SSv2-Temporal, SSv2-Events, and ATP-Hard benchmarks, being a first promising step for using synthetic videos when learning video-language models.
- Architecture, Simulation and Software Stack to Support Post-CMOS Accelerators: The ARCHYTAS Project1
The system architecture and software stack that ARCHYTAS will develop to integrate and support nonconventional hardware accelerators, as well as the simulation software needed for early prototyping of the full system and its components are presented.
- AL-GTD: Deep Active Learning for Gaze Target Detection1
AL-GTD is proposed, an innovative approach that integrates supervised and self-supervised losses within a novel sample acquisition function to perform active learning (AL) and outperforms AL competitors, simultaneously exhibiting superior performance compared to SOTA gaze target detectors when all are trained within a low-data regime.
- Enhancing Robustness of Vision-Language Models through Orthogonality Learning and Self-Regularization1
An orthogonal finetuning method for efficiently updating pretrained weights which enhances robustness and generalization, while a cross-regularization strategy is further exploited to maintain the stability in terms of zero-shot generalization of VLMs, dubbed OrthCR.
- 1
- 1
- 1
- 1
- 1
- Multimodal analysis of free-standing conversational groups1
A multimodal joint head and body pose estimator is described and compared to other recent approaches for head andBody pose estimation and F-formation detection, in particular for the detection of conversational groups or F-formations.
- SVM-based CDMA receiver with incremental active learning1
A sample-by-sample adaptive algorithm for CDMA systems based on incremental SVMs, incorporating an active learning strategy aimed to reduce the complexity of both the training phase and the final classifier is proposed.
- Specificity-aware reinforcement learning for fine-grained open-world classification–
This work proposes a novel specificity-aware reinforcement learning framework, SpeciaRL, to fine-tune reasoning LMMs on fine-grained image classification under the open-world setting, and introduces a dynamic, verifier-based reward signal anchored to the best predictions within online rollouts, promoting specificity while respecting the model's capabilities to prevent incorrect predictions.
- –
- –
- –
- –
- How to Take a Memorable Picture? Empowering Users with Actionable Feedback–
The results indicate that memorability can not only be predicted but also taught and instructed, shifting the focus from mere prediction to actionable feedback for human creators.
- –
- –
- –
- –
- Unlearning Under Imbalance: Benchmarking Fairness in Multimodal LLM Unlearning–
FAIRGET is proposed, the first Visual Question Answering benchmark that evaluates unlearning under unbalanced, realistic, forget requests, and FAUN is proposed, the first unlearning algorithm for MLLMs that forgets unlearning data while preserving model fairness.
- –
- Linear Model Merging Unlocks Simple and Scalable Multimodal Data Mixture Optimization–
This work trains domain-specific multimodal experts and evaluates their weighted parameter-space combinations to estimate the efficacy of corresponding data mixtures, and empirically demonstrates that the merged proxy models exhibit a high rank correlation with models trained on actual data mixtures.
- –
- Training-Free Semantic Multi-Object Tracking with Vision-Language Models–
TF-SMOT is proposed, a training-free SMOT pipeline that composes pretrained components for detection, mask-based tracking, and video-language generation and achieves state-of-the-art tracking performance within the SMOT setting and improves summary and caption quality compared to prior art.
- Zero-Shot Temporal Action Localization Through Textual Guidance–
Results show that, by exploiting rich textual information for improved action localization, TeGu outperforms state-of-the-art ZS-TAL approaches that do not involve training1.
- –
- –
- –
- –
- –
- –
- –
- –
- Recognizing Actions in Videos under Domain Shift–
Some recent works of the research group on UVDA are presented, showing that, thanks to recent advances in deep architectures and to the advent of foundation models, it is possible to deal with more challenging and realistic settings and recognize out-of-distribution classes.
- Introduction to the Special Section on Learning Representations, Similarity, and Associations in Dynamic Multimedia Environments–
The aim of this Special Section on “Learning Representations, Similarities, and Associations in Dynamic Multimedia Environments” is to bring academic researchers and industry developers together for sharing the recent advances and future trends of the representation/similarity learning and association of complex multimedia data.
- –
- –
- –
- –
- Special Issue on Generating Realistic Visual Data of Human Behavior–
This special issue focuses on recent advances and novel methodologies for generating data containing human-like behaviour at multiple scales.
- –
- Introduction to the second workshop on natural language for artificial intelligence–
The goal of the workshop “Natural Language for Artificial Intelligence” (NL4AI) is to provide a meeting forum for stimulating and disseminating research where researchers (especially those affiliated with Italian institutions) can network and discuss their results in an informal way.
- –
- –
- –
- –
- –
- –
- A new algorithm for implementing BSB-based associative memories–
The design of associative memories based on the generalized brain-state-in-a-box (GBSB) neural model is formulated as a set of independent classification tasks, which can be efficiently solved by standard software packages for SVM learning.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-15. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.