Jacopo Cavazza

Giancarlo Paoletti, Jacopo Cavazza, Cigdem Beyan et al.

This paper presents a novel end-to-end method for the problem of skeleton-based unsupervised human action recognition. We propose a new architecture with a convolutional autoencoder that uses graph Laplacian regularization to model the skeletal geometry across the temporal dynamics of actions. Our approach is robust towards viewpoint variations by including a self-supervised gradient reverse layer that ensures generalization across camera views. The proposed method is validated on NTU-60 and NTU-120 large-scale datasets in which it outperforms all prior unsupervised skeleton-based approaches on the cross-subject, cross-view, and cross-setup protocols. Although unsupervised, our learnable representation allows our method even to surpass a few supervised skeleton-based action recognition methods. The code is available in: www.github.com/IIT-PAVIS/UHAR_Skeletal_Laplacian

Shah Nawaz, Jacopo Cavazza, Alessio Del Bue

Zero-shot learning methods rely on fixed visual and semantic embeddings, extracted from independent vision and language models, both pre-trained for other large-scale tasks. This is a weakness of current zero-shot learning frameworks as such disjoint embeddings fail to adequately associate visual and textual information to their shared semantic content. Therefore, we propose to learn semantically grounded and enriched visual information by computing a joint image and text model with a two-stream network on a proxy task. To improve this alignment between image and textual representations, provided by attributes, we leverage ancillary captions to provide grounded semantic information. Our method, dubbed joint embeddings for zero-shot learning is evaluated on several benchmark datasets, improving the performance of existing state-of-the-art methods in both standard ($+1.6$\% on aPY, $+2.6\%$ on FLO) and generalized ($+2.1\%$ on AWA$2$, $+2.2\%$ on CUB) zero-shot recognition.

Federico Marmoreo, Julio Ivan Davila Carrazco, Vittorio Murino et al.

In Generalized Zero-Shot Learning (GZSL), unseen categories (for which no visual data are available at training time) can be predicted by leveraging their class embeddings (e.g., a list of attributes describing them) together with a complementary pool of seen classes (paired with both visual data and class embeddings). Despite GZSL is arguably challenging, we posit that knowing in advance the class embeddings, especially for unseen categories, is an actual limit of the applicability of GZSL towards real-world scenarios. To relax this assumption, we propose Open Zero-Shot Learning (OZSL) to extend GZSL towards the open-world settings. We formalize OZSL as the problem of recognizing seen and unseen classes (as in GZSL) while also rejecting instances from unknown categories, for which neither visual data nor class embeddings are provided. We formalize the OZSL problem introducing evaluation protocols, error metrics and benchmark datasets. We also suggest to tackle the OZSL problem by proposing the idea of performing unknown feature generation (instead of only unseen features generation as done in GZSL). We achieve this by optimizing a generative process to sample unknown class embeddings as complementary to the seen and the unseen. We intend these results to be the ground to foster future research, extending the standard closed-world zero-shot learning (GZSL) with the novel open-world counterpart (OZSL).

Jacopo Cavazza

In Zero-shot learning (ZSL), we classify unseen categories using textual descriptions about their expected appearance when observed (class embeddings) and a disjoint pool of seen classes, for which annotated visual data are accessible. We tackle ZSL by casting a "vanilla" convolutional neural network (e.g. AlexNet, ResNet-101, DenseNet-201 or DarkNet-53) into a zero-shot learner. We do so by crafting the softmax classifier: we freeze its weights using fixed seen classification rules, either semantic (seen class embeddings) or visual (seen class prototypes). Then, we learn a data-driven and ZSL-tailored feature representation on seen classes only to match these fixed classification rules. Given that the latter seamlessly generalize towards unseen classes, while requiring not actual unseen data to be computed, we can perform ZSL inference by augmenting the pool of classification rules at test time while keeping the very same representation we learnt: nowhere re-training or fine-tuning on unseen data is performed. The combination of semantic and visual crafting (by simply averaging softmax scores) improves prior state-of-the-art methods in benchmark datasets for standard, inductive ZSL. After rebalancing predictions to better handle the joint inference over seen and unseen classes, we outperform prior generalized, inductive ZSL methods as well. Also, we gain interpretability at no additional cost, by using neural attention methods (e.g., grad-CAM) as they are. Code will be made publicly available.

Federico Marmoreo, Jacopo Cavazza, Vittorio Murino

In this paper, we address zero-shot learning (ZSL), the problem of recognizing categories for which no labeled visual data are available during training. We focus on the transductive setting, in which unlabelled visual data from unseen classes is available. State-of-the-art paradigms in ZSL typically exploit generative adversarial networks to synthesize visual features from semantic attributes. We posit that the main limitation of these approaches is to adopt a single model to face two problems: 1) generating realistic visual features, and 2) translating semantic attributes into visual cues. Differently, we propose to decouple such tasks, solving them separately. In particular, we train an unconditional generator to solely capture the complexity of the distribution of visual data and we subsequently pair it with a conditional generator devoted to enrich the prior knowledge of the data distribution with the semantic content of the class embeddings. We present a detailed ablation study to dissect the effect of our proposed decoupling approach, while demonstrating its superiority over the related state-of-the-art.

Danilo Greco, Jacopo Cavazza, Alessio Del Bue

Given an unknown audio source, the estimation of time differences-of-arrivals (TDOAs) can be efficiently and robustly solved using blind channel identification and exploiting the cross-correlation identity (CCI). Prior "blind" works have improved the estimate of TDOAs by means of different algorithmic solutions and optimization strategies, while always sticking to the case N = 2 microphones. But what if we can obtain a direct improvement in performance by just increasing N? In this paper we try to investigate this direction, showing that, despite the arguable simplicity, this is capable of (sharply) improving upon state-of-the-art blind channel identification methods based on CCI, without modifying the computational pipeline. Inspired by our results, we seek to warm up the community and the practitioners by paving the way (with two concrete, yet preliminary, examples) towards joint approaches in which advances in the optimization are combined with an increased number of microphones, in order to achieve further improvements.

Giancarlo Paoletti, Jacopo Cavazza, Cigdem Beyan et al.

This paper tackles the problem of human action recognition, defined as classifying which action is displayed in a trimmed sequence, from skeletal data. Albeit state-of-the-art approaches designed for this application are all supervised, in this paper we pursue a more challenging direction: Solving the problem with unsupervised learning. To this end, we propose a novel subspace clustering method, which exploits covariance matrix to enhance the action's discriminability and a timestamp pruning approach that allow us to better handle the temporal dimension of the data. Through a broad experimental validation, we show that our computational pipeline surpasses existing unsupervised approaches but also can result in favorable performances as compared to supervised methods.

Ruggero Ragonesi, Riccardo Volpi, Jacopo Cavazza et al.

We are interested in learning data-driven representations that can generalize well, even when trained on inherently biased data. In particular, we face the case where some attributes (bias) of the data, if learned by the model, can severely compromise its generalization properties. We tackle this problem through the lens of information theory, leveraging recent findings for a differentiable estimation of mutual information. We propose a novel end-to-end optimization strategy, which simultaneously estimates and minimizes the mutual information between the learned representation and the data attributes. When applied on standard benchmarks, our model shows comparable or superior classification performance with respect to state-of-the-art approaches. Moreover, our method is general enough to be applicable to the problem of ``algorithmic fairness'', with competitive results.

Jacopo Cavazza, Pietro Morerio, Vittorio Murino

Despite the recent deep learning (DL) revolution, kernel machines still remain powerful methods for action recognition. DL has brought the use of large datasets and this is typically a problem for kernel approaches, which are not scaling up efficiently due to kernel Gram matrices. Nevertheless, kernel methods are still attractive and more generally applicable since they can equally manage different sizes of the datasets, also in cases where DL techniques show some limitations. This work investigates these issues by proposing an explicit approximated representation that, together with a linear model, is an equivalent, yet scalable, implementation of a kernel machine. Our approximation is directly inspired by the exact feature map that is induced by an RBF Gaussian kernel but, unlike the latter, it is finite dimensional and very compact. We justify the soundness of our idea with a theoretical analysis which proves the unbiasedness of the approximation, and provides a vanishing bound for its variance, which is shown to decrease much rapidly than in alternative methods in the literature. In a broad experimental validation, we assess the superiority of our approximation in terms of 1) ease and speed of training, 2) compactness of the model, and 3) improvements with respect to the state-of-the-art performance.

Pietro Morerio, Jacopo Cavazza, Vittorio Murino

In this work, we face the problem of unsupervised domain adaptation with a novel deep learning approach which leverages on our finding that entropy minimization is induced by the optimal alignment of second order statistics between source and target domains. We formally demonstrate this hypothesis and, aiming at achieving an optimal alignment in practical cases, we adopt a more principled strategy which, differently from the current Euclidean approaches, deploys alignment along geodesics. Our pipeline can be implemented by adding to the standard classification loss (on the labeled source domain), a source-to-target regularizer that is weighted in an unsupervised and data-driven fashion. We provide extensive experiments to assess the superiority of our framework on standard domain and modality adaptation benchmarks.

Jacopo Cavazza, Pietro Morerio, Benjamin Haeffele et al.

Regularization for matrix factorization (MF) and approximation problems has been carried out in many different ways. Due to its popularity in deep learning, dropout has been applied also for this class of problems. Despite its solid empirical performance, the theoretical properties of dropout as a regularizer remain quite elusive for this class of problems. In this paper, we present a theoretical analysis of dropout for MF, where Bernoulli random variables are used to drop columns of the factors. We demonstrate the equivalence between dropout and a fully deterministic model for MF in which the factors are regularized by the sum of the product of squared Euclidean norms of the columns. Additionally, we inspect the case of a variable sized factorization and we prove that dropout achieves the global minimum of a convex approximation problem with (squared) nuclear norm regularization. As a result, we conclude that dropout can be used as a low-rank regularizer with data dependent singular-value thresholding.

Jacopo Cavazza, Connor Lane, Benjamin D. Haeffele et al.

Dropout is a simple yet effective algorithm for regularizing neural networks by randomly dropping out units through Bernoulli multiplicative noise, and for some restricted problem classes, such as linear or logistic regression, several theoretical studies have demonstrated the equivalence between dropout and a fully deterministic optimization problem with data-dependent Tikhonov regularization. This work presents a theoretical analysis of dropout for matrix factorization, where Bernoulli random variables are used to drop a factor, thereby attempting to control the size of the factorization. While recent work has demonstrated the empirical effectiveness of dropout for matrix factorization, a theoretical understanding of the regularization properties of dropout in this context remains elusive. This work demonstrates the equivalence between dropout and a fully deterministic model for matrix factorization in which the factors are regularized by the sum of the product of the norms of the columns. While the resulting regularizer is closely related to a variational form of the nuclear norm, suggesting that dropout may limit the size of the factorization, we show that it is possible to trivially lower the objective value by doubling the size of the factorization. We show that this problem is caused by the use of a fixed dropout rate, which motivates the use of a rate that increases with the size of the factorization. Synthetic experiments validate our theoretical findings.

Jacopo Cavazza, Pietro Morerio, Vittorio Murino

3D action recognition was shown to benefit from a covariance representation of the input data (joint 3D positions). A kernel machine feed with such feature is an effective paradigm for 3D action recognition, yielding state-of-the-art results. Yet, the whole framework is affected by the well-known scalability issue. In fact, in general, the kernel function has to be evaluated for all pairs of instances inducing a Gram matrix whose complexity is quadratic in the number of samples. In this work we reduce such complexity to be linear by proposing a novel and explicit feature map to approximate the kernel function. This allows to train a linear classifier with an explicit feature encoding, which implicitly implements a Log-Euclidean machine in a scalable fashion. Not only we prove that the proposed approximation is unbiased, but also we work out an explicit strong bound for its variance, attesting a theoretical superiority of our approach with respect to existing ones. Experimentally, we verify that our representation provides a compact encoding and outperforms other approximation schemes on a number of publicly available benchmark datasets for 3D action recognition.

Andrea Zunino, Jacopo Cavazza, Atesh Koul et al.

In computer vision, video-based approaches have been widely explored for the early classification and the prediction of actions or activities. However, it remains unclear whether this modality (as compared to 3D kinematics) can still be reliable for the prediction of human intentions, defined as the overarching goal embedded in an action sequence. Since the same action can be performed with different intentions, this problem is more challenging but yet affordable as proved by quantitative cognitive studies which exploit the 3D kinematics acquired through motion capture systems. In this paper, we bridge cognitive and computer vision studies, by demonstrating the effectiveness of video-based approaches for the prediction of human intentions. Precisely, we propose Intention from Motion, a new paradigm where, without using any contextual information, we consider instantaneous grasping motor acts involving a bottle in order to forecast why the bottle itself has been reached (to pass it or to place in a box, or to pour or to drink the liquid inside). We process only the grasping onsets casting intention prediction as a classification framework. Leveraging on our multimodal acquisition (3D motion capture data and 2D optical videos), we compare the most commonly used 3D descriptors from cognitive studies with state-of-the-art video-based techniques. Since the two analyses achieve an equivalent performance, we demonstrate that computer vision tools are effective in capturing the kinematics and facing the cognitive problem of human intention prediction.

Jacopo Cavazza, Pietro Morerio, Vittorio Murino

Human action recognition from skeletal data is a hot research topic and important in many open domain applications of computer vision, thanks to recently introduced 3D sensors. In the literature, naive methods simply transfer off-the-shelf techniques from video to the skeletal representation. However, the current state-of-the-art is contended between to different paradigms: kernel-based methods and feature learning with (recurrent) neural networks. Both approaches show strong performances, yet they exhibit heavy, but complementary, drawbacks. Motivated by this fact, our work aims at combining together the best of the two paradigms, by proposing an approach where a shallow network is fed with a covariance representation. Our intuition is that, as long as the dynamics is effectively modeled, there is no need for the classification network to be deep nor recurrent in order to score favorably. We validate this hypothesis in a broad experimental analysis over 6 publicly available datasets.

Pietro Morerio, Jacopo Cavazza, Riccardo Volpi et al.

Dropout is a very effective way of regularizing neural networks. Stochastically "dropping out" units with a certain probability discourages over-specific co-adaptations of feature detectors, preventing overfitting and improving network generalization. Besides, Dropout can be interpreted as an approximate model aggregation technique, where an exponential number of smaller networks are averaged in order to get a more powerful ensemble. In this paper, we show that using a fixed dropout probability during training is a suboptimal choice. We thus propose a time scheduling for the probability of retaining neurons in the network. This induces an adaptive regularization scheme that smoothly increases the difficulty of the optimization problem. This idea of "starting easy" and adaptively increasing the difficulty of the learning problem has its roots in curriculum learning and allows one to train better models. Indeed, we prove that our optimization strategy implements a very general curriculum scheme, by gradually adding noise to both the input and intermediate feature representations within the network architecture. Experiments on seven image classification datasets and different network architectures show that our method, named Curriculum Dropout, frequently yields to better generalization and, at worst, performs just as well as the standard Dropout method.

Jacopo Cavazza, Vittorio Murino

This paper addresses the scalar regression problem through a novel solution to exactly optimize the Huber loss in a general semi-supervised setting, which combines multi-view learning and manifold regularization. We propose a principled algorithm to 1) avoid computationally expensive iterative schemes while 2) adapting the Huber loss threshold in a data-driven fashion and 3) actively balancing the use of labelled data to remove noisy or inconsistent annotations at the training stage. In a wide experimental evaluation, dealing with diverse applications, we assess the superiority of our paradigm which is able to combine robustness towards noise with both strong performance and low computational cost.

Andrea Zunino, Jacopo Cavazza, Atesh Koul et al.

In this paper, we address the new problem of the prediction of human intents. There is neuro-psychological evidence that actions performed by humans are anticipated by peculiar motor acts which are discriminant of the type of action going to be performed afterwards. In other words, an actual intent can be forecast by looking at the kinematics of the immediately preceding movement. To prove it in a computational and quantitative manner, we devise a new experimental setup where, without using contextual information, we predict human intents all originating from the same motor act. We posit the problem as a classification task and we introduce a new multi-modal dataset consisting of a set of motion capture marker 3D data and 2D video sequences, where, by only analysing very similar movements in both training and test phases, we are able to predict the underlying intent, i.e., the future, never observed action. We also present an extensive experimental evaluation as a baseline, customizing state-of-the-art techniques for either 3D and 2D data analysis. Realizing that video processing methods lead to inferior performance but show complementary information with respect to 3D data sequences, we developed a 2D+3D fusion analysis where we achieve better classification accuracies, attesting the superiority of the multimodal approach for the context-free prediction of human intents.

Andrea Zunino, Jacopo Cavazza, Vittorio Murino

By thoroughly revisiting the classic human action recognition paradigm, this paper aims at proposing a new approach for the design of effective action classification systems. Taking as testbed publicly available three-dimensional (MoCap) action/activity datasets, we analyzed and validated different training/testing strategies. In particular, considering that each human action in the datasets is performed several times by different subjects, we were able to precisely quantify the effect of inter- and intra-subject variability, so as to figure out the impact of several learning approaches in terms of classification performance. The net result is that standard testing strategies consisting in cross-validating the algorithm using typical splits of the data (holdout, k-fold, or one-subject-out) is always outperformed by a "personalization" strategy which learns how a subject is performing an action. In other words, it is advantageous to customize (i.e., personalize) the method to learn the actions carried out by each subject, rather than trying to generalize the actions executions across subjects. Consequently, we finally propose an action recognition framework consisting of a two-stage classification approach where, given a test action, the subject is first identified before the actual recognition of the action takes place. Despite the basic, off-the-shelf descriptors and standard classifiers adopted, we noted a relevant increase in performance with respect to standard state-of-the-art algorithms, so motivating the usage of personalized approaches for designing effective action recognition systems.

Jacopo Cavazza, Andrea Zunino, Marco San Biagio et al.

In this paper we aim at increasing the descriptive power of the covariance matrix, limited in capturing linear mutual dependencies between variables only. We present a rigorous and principled mathematical pipeline to recover the kernel trick for computing the covariance matrix, enhancing it to model more complex, non-linear relationships conveyed by the raw data. To this end, we propose Kernelized-COV, which generalizes the original covariance representation without compromising the efficiency of the computation. In the experiments, we validate the proposed framework against many previous approaches in the literature, scoring on par or superior with respect to the state of the art on benchmark datasets for 3D action recognition.