Human Mesh Recovery (HMR) from a single RGB image is a highly ambiguous problem, as an infinite set of 3D interpretations can explain the 2D observation equally well. Nevertheless, most HMR methods overlook this issue and make a single prediction without accounting for this ambiguity. A few approaches generate a distribution of human meshes, enabling the sampling of multiple predictions; however, none of them is competitive with the latest single-output model when making a single prediction. This work proposes a new approach based on masked generative modeling. By tokenizing the human pose and shape, we formulate the HMR task as generating a sequence of discrete tokens conditioned on an input image. We introduce MEGA, a MaskEd Generative Autoencoder trained to recover human meshes from images and partial human mesh token sequences. Given an image, our flexible generation scheme allows us to predict a single human mesh in deterministic mode or to generate multiple human meshes in stochastic mode. Experiments on in-the-wild benchmarks show that MEGA achieves state-of-the-art performance in deterministic and stochastic modes, outperforming single-output and multi-output approaches. See the project page at https://g-fiche.github.io/research-pages/mega/.

This paper proposes a new unsupervised audiovisual speech enhancement (AVSE) approach that combines a diffusion-based audio-visual speech generative model with a non-negative matrix factorization (NMF) noise model. First, the diffusion model is pre-trained on clean speech conditioned on corresponding video data to simulate the speech generative distribution. This pre-trained model is then paired with the NMF-based noise model to estimate clean speech iteratively. Specifically, a diffusion-based posterior sampling approach is implemented within the reverse diffusion process, where after each iteration, a speech estimate is obtained and used to update the noise parameters. Experimental results confirm that the proposed AVSE approach not only outperforms its audio-only counterpart but also generalizes better than a recent supervised-generative AVSE method. Additionally, the new inference algorithm offers a better balance between inference speed and performance compared to the previous diffusion-based method. Code and demo available at: https://jeaneudesayilo.github.io/fast_UdiffSE

This article introduces AnCoGen, a novel method that leverages a masked autoencoder to unify the analysis, control, and generation of speech signals within a single model. AnCoGen can analyze speech by estimating key attributes, such as speaker identity, pitch, content, loudness, signal-to-noise ratio, and clarity index. In addition, it can generate speech from these attributes and allow precise control of the synthesized speech by modifying them. Extensive experiments demonstrated the effectiveness of AnCoGen across speech analysis-resynthesis, pitch estimation, pitch modification, and speech enhancement. Code and audio examples are available online1.

We explore unsupervised speech enhancement using diffusion models as expressive generative priors for clean speech. Existing approaches guide the reverse diffusion process using noisy speech through an approximate, noise-perturbed likelihood score, combined with the unconditional score via a trade-off hyperparameter. In this work, we propose two alternative algorithms that directly model the conditional reverse transition distribution of diffusion states. The first method integrates the diffusion prior with the observation model in a principled way, removing the need for hyperparameter tuning. The second defines a diffusion process over the noisy speech itself, yielding a fully tractable and exact likelihood score. Experiments on the WSJ0-QUT and VoiceBank-DEMAND datasets demonstrate improved enhancement metrics and greater robustness to domain shifts compared to both supervised and unsupervised baselines.

In this work, we address the task of unconditional head motion generation to animate still human faces in a low-dimensional semantic space from a single reference pose. Different from traditional audio-conditioned talking head generation that seldom puts emphasis on realistic head motions, we devise a GAN-based architecture that learns to synthesize rich head motion sequences over long duration while maintaining low error-accumulation levels. In particular, the autoregressive generation of incremental outputs ensures smooth trajectories, while a multi-scale discriminator on input pairs drives generation toward better handling of high- and low-frequency signals and less mode collapse. We experimentally demonstrate the relevance of the proposed method and show its superiority compared to models that attained state-of-the-art performances on similar tasks.

We address speech enhancement based on variational autoencoders, which involves learning a speech prior distribution in the time-frequency (TF) domain. A zero-mean complex-valued Gaussian distribution is usually assumed for the generative model, where the speech information is encoded in the variance as a function of a latent variable. In contrast to this commonly used approach, we propose a weighted variance generative model, where the contribution of each spectrogram time-frame in parameter learning is weighted. We impose a Gamma prior distribution on the weights, which would effectively lead to a Student’s t-distribution instead of Gaussian for speech generative modeling. We develop efficient training and speech enhancement algorithms based on the proposed generative model. Our experimental results on spectrogram auto-encoding and speech enhancement demonstrate the effectiveness and robustness of the proposed approach compared to the standard unweighted variance model.

High-dimensional data such as natural images or speech signals exhibit some form of regularity, preventing their dimensions from varying independently. This suggests that there exists a lower dimensional latent representation from which the high-dimensional observed data were generated. Uncovering the hidden explanatory features of complex data is the goal of representation learning, and deep latent variable generative models have emerged as promising unsupervised approaches. In particular, the variational autoencoder (VAE) which is equipped with both a generative and an inference model allows for the analysis, transformation, and generation of various types of data. Over the past few years, the VAE has been extended to deal with data that are either multimodal or dynamical (i.e., sequential). In this paper, we present a multimodal and dynamical VAE (MDVAE) applied to unsupervised audiovisual speech representation learning. The latent space is structured to dissociate the latent dynamical factors that are shared between the modalities from those that are specific to each modality. A static latent variable is also introduced to encode the information that is constant over time within an audiovisual speech sequence. The model is trained in an unsupervised manner on an audiovisual emotional speech dataset, in two stages. In the first stage, a vector quantized VAE (VQ-VAE) is learned independently for each modality, without temporal modeling. The second stage consists in learning the MDVAE model on the intermediate representation of the VQ-VAEs before quantization. The disentanglement between static versus dynamical and modality-specific versus modality-common information occurs during this second training stage. Extensive experiments are conducted to investigate how audiovisual speech latent factors are encoded in the latent space of MDVAE. These experiments include manipulating audiovisual speech, audiovisual facial image denoising, and audiovisual speech emotion recognition. The results show that MDVAE effectively combines the audio and visual information in its latent space. They also show that the learned static representation of audiovisual speech can be used for emotion recognition with few labeled data, and with better accuracy compared with unimodal baselines and a state-of-the-art supervised model based on an audiovisual transformer architecture.

By observing a scene and listening to corresponding audio cues, humans can easily recognize where the sound is. To achieve such cross-modal perception on machines, existing methods take advantage of the maps obtained by interpolation operations to localize the sound source. As semantic object-level localization is more attractive for prospective practical applications, we argue that these map-based methods only offer a coarse-grained and indirect description of the sound source. Additionally, these methods utilize a single audio-visual tuple at a time during self-supervised learning, causing the model to lose the crucial chance to reason about the data distribution of large-scale audio-visual samples. Although the introduction of Audio-Visual Contrastive Learning (AVCL) can effectively alleviate this issue, the contrastive set constructed by randomly sampling is based on the assumption that the audio and visual segments from all other videos are not semantically related. Since the resulting contrastive set contains a large number of faulty negatives, we believe that this assumption is rough. In this paper, we advocate a novel proposal-based solution that directly localizes the semantic object-level sound source, without any manual annotations. The Global Response Map (GRM) is incorporated as an unsupervised spatial constraint to filter those instances corresponding to a large number of sound-unrelated regions. As a result, our proposal-based Sound Source Localization (SSL) can be cast into a simpler Multiple Instance Learning (MIL) problem. To overcome the limitation of random sampling in AVCL, we propose a novel Active Contrastive Set Mining (ACSM) to mine the contrastive sets with informative and diverse negatives for robust AVCL. Our approaches achieve state-of-the-art (SOTA) performance when compared to several baselines on multiple SSL datasets with diverse scenarios.

Self-supervised learning models have been shown to learn rich visual representations without requiring human annotations. However, in many real-world scenarios, labels are partially available, motivating a recent line of work on semi-supervised methods inspired by self-supervised principles. In this paper, we propose a conceptually simple yet empirically powerful approach to turn clustering-based self-supervised methods such as SwAV or DINO into semi-supervised learners. More precisely, we introduce 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. This approach may be interpreted as imposing the cluster centroids to be class prototypes. Despite its simplicity, we provide empirical evidence that our approach is highly effective and achieves state-of-the-art performance on CI-FAR100 and ImageNet.

The dynamical variational autoencoders (DVAEs) are a family of latent-variable deep generative models that extends the VAE to model a sequence of observed data and a corresponding sequence of latent vectors. In almost all the DVAEs of the literature, the temporal dependencies within each sequence and across the two sequences are modeled with recurrent neural networks. In this paper, we propose to model speech signals with the Hierarchical Transformer DVAE (HiT-DVAE), which is a DVAE with two levels of latent variable (sequence-wise and frame-wise) and in which the temporal dependencies are implemented with the Transformer architecture. We show that HiT-DVAE outperforms several other DVAEs for speech spectrogram modeling, while enabling a simpler training procedure, revealing its high potential for downstream low-level speech processing tasks such as speech enhancement.

This work builds on a previous work on unsupervised speech enhancement using a dynamical variational autoencoder (DVAE) as the clean speech model and non-negative matrix factorization (NMF) as the noise model. We propose to replace the NMF noise model with a deep dynamical generative model (DDGM) depending either on the DVAE latent variables, or on the noisy observations, or on both. This DDGM can be trained in three configurations: noise-agnostic, noise-dependent and noise adaptation after noise-dependent training. Experimental results show that the proposed method achieves competitive performance compared to state-of-the-art unsupervised speech enhancement methods, while the noise-dependent training configuration yields a much more time-efficient inference process.

Pose and motion priors are crucial for recovering realistic and accurate human motion from noisy observations. Substantial progress has been made on pose and shape estimation from images, and recent works showed impressive results using priors to refine frame-wise predictions. However, a lot of motion priors only model transitions between consecutive poses and are used in time-consuming optimization procedures, which is problematic for many applications requiring real-time motion capture. We introduce Motion-DVAE, a motion prior to capture the short-term dependencies of human motion. As part of the dynamical variational autoencoder (DVAE) models family, Motion-DVAE combines the generative capability of VAE models and the temporal modeling of recurrent architectures. Together with Motion-DVAE, we introduce an unsupervised learned denoising method unifying regression- and optimization-based approaches in a single framework for real-time 3D human pose estimation. Experiments show that the proposed approach reaches competitive performance with state-of-the-art methods while being much faster.

This paper tackles the problem of human motion prediction, consisting in forecasting future body poses from historically observed sequences. State-of-the-art approaches provide good results, however, they rely on deep learning architectures of arbitrary complexity, such as Recurrent Neural Networks (RNN), Transformers or Graph Convolutional Networks (GCN), typically requiring multiple training stages and more than 2 million parameters. In this paper, we show that, after combining with a series of standard practices, such as applying Discrete Cosine Transform (DCT), predicting residual displacement of joints and optimizing velocity as an auxiliary loss, a light-weight network based on multi-layer perceptrons (MLPs) with only 0.14 million parameters can surpass the state-of-the-art performance. An exhaustive evaluation on the Human3.6M, AMASS, and 3DPW datasets shows that our method, named siMLpe, consistently outperforms all other approaches. We hope that our simple method could serve as a strong baseline for the community and allow re-thinking of the human motion prediction problem. The code is publicly available at https://github.com/dulucas/siMLPe.

Transformers have proven superior performance for a wide variety of tasks since they were introduced. In recent years, they have drawn attention from the vision community in tasks such as image classification and object detection. Despite this wave, an accurate and efficient multiple-object tracking (MOT) method based on transformers is yet to be designed. We argue that the direct application of a transformer architecture with quadratic complexity and insufficient noise-initialized sparse queries - is not optimal for MOT. We propose TransCenter, a transformer-based MOT architecture with dense representations for accurately tracking all the objects while keeping a reasonable runtime. Methodologically, we propose the use of image-related dense detection queries and efficient sparse tracking queries produced by our carefully designed query learning networks (QLN). On one hand, the dense image-related detection queries allow us to infer targets' locations globally and robustly through dense heatmap outputs. On the other hand, the set of sparse tracking queries efficiently interacts with image features in our TransCenter Decoder to associate object positions through time. As a result, TransCenterexhibits remarkable performance improvements and outperforms by a large margin the current state-of-the-art methods in two standard MOT benchmarks with two tracking settings (public/private). TransCenter is also proven efficient and accurate by an extensive ablation study and, comparisons to more naive alternatives and concurrent works. The code is made publicly available at https://github.com/yihongxu/transcenter.

A fundamental and challenging problem in deep learning is catastrophic forgetting, i.e., the tendency of neural networks to fail to preserve the knowledge acquired from old tasks when learning new tasks. This problem has been widely investigated in the research community and several Incremental Learning (IL) approaches have been proposed in the past years. While earlier works in computer vision have mostly focused on image classification and object detection, more recently some IL approaches for semantic segmentation have been introduced. These previous works showed that, despite its simplicity, knowledge distillation can be effectively employed to alleviate catastrophic forgetting. In this paper, we follow this research direction and, inspired by recent literature on contrastive learning, we propose a novel distillation framework, Uncertainty-aware Contrastive Distillation (UCD). In a nutshell, UCDis operated by introducing a novel distillation loss 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. In order to mitigate catastrophic forgetting, we contrast features of the new model with features extracted by a frozen model learned at the previous incremental step. Our experimental results demonstrate the advantage of the proposed distillation technique, which can be used in synergy with previous IL approaches, and leads to state-of-art performance on three commonly adopted benchmarks for incremental semantic segmentation.

National audience

Prediction of human actions in social interactions has important applications in the design of social robots or artificial avatars. In this paper, we focus on a unimodal representation of interactions and propose to tackle interaction generation in a data-driven fashion. In particular, we model human interaction generation as a discrete multi-sequence generation problem and present SocialInteractionGAN, a novel adversarial architecture for conditional interaction generation. Our model builds on a recurrent encoder-decoder generator network and a dual-stream discriminator, that jointly evaluates the realism of interactions and individual action sequences and operates at different time scales. Crucially, contextual information on interacting participants is shared among agents and reinjected in both the generation and the discriminator evaluation processes. Experiments show that albeit dealing with low dimensional data, SocialInteractionGAN succeeds in producing high realism action sequences of interacting people, comparing favorably to a diversity of recurrent and convolutional discriminator baselines, and we argue that this work will constitute a first stone towards higher dimensional and multimodal interaction generation. Evaluations are conducted using classical GAN metrics, that we specifically adapt for discrete sequential data. Our model is shown to properly learn the dynamics of interaction sequences, while exploiting the full range of available actions.

International audience

Self-supervised models have been shown to produce comparable or better visual representations than their su-pervised counterparts when trained offline on unlabeled data at scale. However, their efficacy is catastrophically reduced in a Continual Learning (CL) scenario where data is presented to the model sequentially. In this paper, we show 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. This enables us to devise a framework for Continual self-supervised visual representation Learning that (i) significantly improves the quality of the learned representations, (ii) is compatible with several state-of-the-art self-supervised objectives, and (iii) needs little to no hyperparameter tuning. We demonstrate the ef-fectiveness of our approach empirically by training six pop-ular self-supervised models in various CL settings. Code: github.com/DonkeyShot21/cassle.

Human motion prediction aims to forecast future poses given a sequence of past 3D skeletons. While this problem has recently received increasing attention, it has mostly been tackled for single humans in isolation. In this paper, we explore this problem when dealing with humans performing collaborative tasks, we seek to predict the future motion of two interacted persons given two sequences of their past skeletons. We propose a novel cross interaction attention mechanism that exploits historical information of both persons, and learns to predict cross dependencies between the two pose sequences. Since no dataset to train such interactive situations is available, we collected ExPI (Extreme Pose Interaction) dataset, a new lab-based per-son interaction dataset of professional dancers performing Lindy-hop dancing actions, which contains 115 sequences with 30K frames annotated with 3D body poses and shapes. We thoroughly evaluate our cross interaction network on ExPI and show that both in short- and long-term predictions, it consistently outperforms state-of-the-art methods for single-person motion prediction. Our code and dataset are available at: https://team.inria.fr/robotlearn/multi-person-extreme-motionprediction/

Humans can easily recognize where and how the sound is produced via watching a scene and listening to corresponding audio cues. To achieve such cross-modal perception on machines, existing methods only use the maps generated by interpolation operations to localize the sound source. As semantic object-level localization is more attractive for potential practical applications, we argue that these existing map-based approaches only provide a coarse-grained and indirect description of the sound source. In this pa-per, we advocate a novel proposal-based paradigm that can directly perform semantic object-level localization, without any manual annotations. We incorporate the global re-sponse map as an unsupervised spatial constraint to weight the proposals according to how well they cover the esti-mated global shape of the sound source. As a result, our proposal-based sound source localization can be cast into a simpler Multiple Instance Learning (MIL) problem by filtering those instances corresponding to large sound-unrelated regions. Our method achieves state-of-the-art (SOTA) per-formance when compared to several baselines on multiple datasets.

This paper investigates the impact of head movements on audio-visual speech enhancement (AVSE). Although being a common conversational feature, head movements have been ignored by past and recent studies: they challenge today’s learning-based methods as they often degrade the performance of models that are trained on clean, frontal, and steady face images. To alleviate this problem, we propose to use robust face frontalization (RFF) in combination with an AVSE method based on a variational auto-encoder (VAE) model. We briefly describe the basic ingredients of the proposed pipeline and we perform experiments with a recently released audio-visual dataset. In the light of these experiments, and based on three standard metrics, namely STOI, PESQ and SI-SDR, we conclude that RFF improves the performance of AVSE by a considerable margin.1

The recent success of audio-visual representation learning can be largely attributed to their pervasive property of audio-visual synchronization, which can be used as self-annotated supervision. As a state-of-the-art solution, Audio-Visual Instance Discrimination (AVID) extends instance discrimination to the audio-visual realm. Existing AVID methods construct the contrastive set by random sampling based on the assumption that the audio and visual clips from all other videos are not semantically related. We argue that this assumption is rough, since the resulting contrastive sets have a large number of faulty negatives. In this paper, we overcome this limitation by proposing a novel Active Contrastive Set Mining (ACSM) that aims to mine the contrastive sets with informative and diverse negatives for robust AVID. Moreover, we also integrate a semantically-aware hard-sample mining strategy into our ACSM. The proposed ACSM is implemented into two most recent state-of-the-art AVID methods and significantly improves their performance. Extensive experiments conducted on both action and sound recognition on multiple datasets show the remarkably improved performance of our method.

Transversely to all multimedia research, the methods and tools are often shared by several MM topics and applications. M4MM aims to foster discussion around fundamental methodologies and tools that are used in various multimedia research topics. To that aim, the technical program of the workshop consists of two keynote speakers, on complementary and very relevant methods and tools for MM, as well as four technical papers. The complete M4MM'22 workshop proceedings are available at: https://dl.acm.org/doi/proceedings/10.1145/3552487

Multi-scale representations deeply learned via convolutional neural networks have shown tremendous importance for various pixel-level prediction problems. In this paper we present a novel approach that advances the state of the art on pixel-level prediction in a fundamental aspect, i.e. structured multi-scale features learning and fusion. In contrast to previous works directly considering multi-scale feature maps obtained from the inner layers of a primary CNN architecture, and simply fusing the features with weighted averaging or concatenation, we propose a probabilistic graph attention network structure based on a novel Attention-Gated Conditional Random Fields (AG-CRFs) model for learning and fusing multi-scale representations in a principled manner. In order to further improve the learning capacity of the network structure, we propose to exploit feature dependant conditional kernels within the deep probabilistic framework. Extensive experiments are conducted on four publicly available datasets (i.e. BSDS500, NYUD-V2, KITTI and Pascal-Context) and on three challenging pixel-wise prediction problems involving both discrete and continuous labels (i.e. monocular depth estimation, object contour prediction and semantic segmentation). 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.

Dynamical variational autoencoders (DVAEs) are a class of deep generative models with latent variables, dedicated to model time series of high-dimensional data. DVAEs can be considered as extensions of the variational autoencoder (VAE) that include temporal dependencies between successive observed and/or latent vectors. Previous work has shown the interest of using DVAEs over the VAE for speech spectrograms modeling. Independently, the VAE has been successfully applied to speech enhancement in noise, in an unsupervised noise-agnostic set-up that requires neither noise samples nor noisy speech samples at training time, but only requires clean speech signals. In this paper, we extend these works to DVAE-based single-channel unsupervised speech enhancement, hence exploiting both speech signals unsupervised representation learning and dynamics modeling. We propose an unsupervised speech enhancement algorithm that combines a DVAE speech prior pre-trained on clean speech signals with a noise model based on nonnegative matrix factorization, and we derive a variational expectation-maximization (VEM) algorithm to perform speech enhancement. The algorithm is presented with the most general DVAE formulation and is then applied with three specific DVAE models to illustrate the versatility of the framework. Experimental results show that the proposed DVAE-based approach outperforms its VAE-based counterpart, as well as several supervised and unsupervised noise-dependent baselines, especially when the noise type is unseen during training.

Modeling the temporal behavior of data is of primordial importance in many scientific and engineering fields. Baseline methods assume that both the dynamic and observation equations follow linear-Gaussian models. However, there are many real-world processes that cannot be characterized by a single linear behavior. Alternatively, it is possible to consider a piecewise linear model which, combined with a switching mechanism, is well suited when several modes of behavior are needed. Nevertheless, switching dynamical systems are intractable because their computational complexity increases exponentially with time. In this article, we propose a variational approximation of piecewise linear dynamical systems. We provide full details of the derivation of two variational expectation-maximization algorithms: a filter and a smoother. We show that the model parameters can be split into two sets: static and dynamic parameters, and that the former parameters can be estimated offline together with the number of linear modes, or the number of states of the switching variable. We apply the proposed method to the head-pose tracking, and we thoroughly compare our algorithms with several state of the art trackers.

Recently, audio-visual speech enhancement has been tackled in the unsupervised settings based on variational auto-encoders (VAEs), where during training only clean data is used to train a generative model for speech, which at test time is combined with a noise model, e.g. nonnegative matrix factorization (NMF), whose parameters are learned without supervision. Consequently, the proposed model is agnostic to the noise type. When visual data are clean, audio-visual VAE-based architectures usually outperform the audio-only counterpart. The opposite happens when the visual data are corrupted by clutter, e.g. the speaker not facing the camera. In this paper, we propose to find the optimal combination of these two architectures through time. More precisely, we introduce the use of a latent sequential variable with Markovian dependencies to switch between different VAE architectures through time in an unsupervised manner: leading to switching variational auto-encoder (SwVAE). We propose a variational factorization to approximate the computationally intractable posterior distribution. We also derive the corresponding variational expectation-maximization algorithm to estimate the parameters of the model and enhance the speech signal. Our experiments demonstrate the promising performance of SwVAE.

Accepted to Interspeech 2021. arXiv admin note: text overlap with arXiv:2008.12595

Recent literature addressed the monocular 3D pose estimation task very satisfactorily. In these studies, different persons are usually treated as independent pose instances to estimate. However, in many every-day situations, people are interacting, and the pose of an individual depends on the pose of his/her interactees. In this paper, we investigate how to exploit this dependency to enhance current - and possibly future - deep networks for 3D monocular pose estimation. Our pose interacting network, or PI-Net, inputs the initial pose estimates of a variable number of interactees into a recurrent architecture used to refine the pose of the person-of-interest. Evaluating such a method is challenging due to the limited availability of public annotated multi-person 3D human pose datasets. We demonstrate the effectiveness of our method in the MuPoTS dataset, setting the new state-of-the-art on it. Qualitative results on other multi-person datasets (for which 3D pose ground-truth is not available) showcase the proposed PI-Net. PI-Net is implemented in PyTorch and the code will be made available upon acceptance of the paper.

In this article, we address the problem of tracking multiple speakers via the fusion of visual and auditory information. We propose to exploit the complementary nature and roles of these two modalities in order to accurately estimate smooth trajectories of the tracked persons, to deal with the partial or total absence of one of the modalities over short periods of time, and to estimate the acoustic status-either speaking or silent-of each tracked person over time. We propose to cast the problem at hand into a generative audio-visual fusion (or association) model formulated as a latent-variable temporal graphical model. This may well be viewed as the problem of maximizing the posterior joint distribution of a set of continuous and discrete latent variables given the past and current observations, which is intractable. We propose a variational inference model which amounts to approximate the joint distribution with a factorized distribution. The solution takes the form of a closed-form expectation maximization procedure. We describe in detail the inference algorithm, we evaluate its performance and we compare it with several baseline methods. These experiments show that the proposed audio-visual tracker performs well in informal meetings involving a time-varying number of people.

The recent trend in vision-based multi-object tracking (MOT) is heading towards leveraging the representational power of deep learning to jointly learn to detect and track objects. However, existing methods train only certain sub-modules using loss functions that often do not correlate with established tracking evaluation measures such as Multi-Object Tracking Accuracy (MOTA) and Precision (MOTP). As these measures are not differentiable, the choice of appropriate loss functions for end-to-end training of multi-object tracking methods is still an open research problem. In this paper, we bridge this gap by proposing a differentiable proxy of MOTA and MOTP, which we combine in a loss function suitable for end-to-end training of deep multi-object trackers. As a key ingredient, we propose a Deep Hungarian Net (DHN) module that approximates the Hungarian matching algorithm. DHN allows estimating the correspondence between object tracks and ground truth objects to compute differentiable proxies of MOTA and MOTP, which are in turn used to optimize deep trackers directly. We experimentally demonstrate that the proposed differentiable framework improves the performance of existing multi-object trackers, and we establish a new state of the art on the MOTChallenge benchmark. Our code is publicly available from https://github.com/yihongXU/deepMOT.

This paper presents a generative approach to speech enhancement based on a recurrent variational autoencoder (RVAE). The deep generative speech model is trained using clean speech signals only, and it is combined with a nonnegative matrix factorization noise model for speech enhancement. We propose a variational expectation-maximization algorithm where the encoder of the RVAE is finetuned at test time, to approximate the distribution of the latent variables given the noisy speech observations. Compared with previous approaches based on feed-forward fully-connected architectures, the proposed recurrent deep generative speech model induces a posterior temporal dynamic over the latent variables, which is shown to improve the speech enhancement results.

Recently, an audio-visual speech generative model based on variational autoencoder (VAE) has been proposed, which is combined with a nonnegative matrix factorization (NMF) model for noise variance to perform unsupervised speech enhancement. When visual data is clean, speech enhancement with audio-visual VAE shows a better performance than with audio-only VAE, which is trained on audio-only data. However, audio-visual VAE is not robust against noisy visual data, e.g., when for some video frames, speaker face is not frontal or lips region is occluded. In this paper, we propose a robust unsupervised audio-visual speech enhancement method based on a per-frame VAE mixture model. This mixture model consists of a trained audio-only VAE and a trained audio-visual VAE. The motivation is to skip noisy visual frames by switching to the audio-only VAE model. We present a variational expectation-maximization method to estimate the parameters of the model. Experiments show the promising performance of the proposed method.

Unsupervised person re-ID is the task of identifying people on a target data set for which the ID labels are unavailable during training. In this paper, we propose to unify two trends in unsupervised person re-ID: clustering & fine-tuning and adversarial learning. On one side, clustering groups training images into pseudo-ID labels, and uses them to fine-tune the feature extractor. On the other side, adversarial learning is used, inspired by domain adaptation, to match distributions from different domains. Since target data is distributed across different camera viewpoints, we propose to model each camera as an independent domain, and aim to learn domain-independent features. Straightforward adversarial learning yields negative transfer, we thus introduce a conditioning vector to mitigate this undesirable effect. In our framework, the centroid of the cluster to which the visual sample belongs is used as conditioning vector of our conditional adversarial network, where the vector is permutation invariant (clusters ordering does not matter) and its size is independent of the number of clusters. To our knowledge, we are the first to propose the use of conditional adversarial networks for unsupervised person re-ID. We evaluate the proposed architecture on top of two state-of-the-art clustering-based unsupervised person re-identification (re-ID) methods on four different experimental settings with three different data sets and set the new state-of-the-art performance on all four of them. Our code and model is publicly available at https://team.inria.fr/perception/research/canu-reid/.

The series of FAT/FAccT events aim at bringing together researchers and practitioners interested in fairness, accountability, transparency and ethics of computational methods. The FATE/MM workshop focuses on addressing these issues in the Multimedia field. Multimedia computing technologies operate today at an unprecedented scale, with a growing community of scientists interested in multimedia models, tools and applications. Such continued growth has great implications not only for the scientific community, but also for the society as a whole. Typical risks of large-scale computational models include model bias and algorithmic discrimination. These risks become particularly prominent in the multimedia field, which historically has been focusing on user-centered technologies. To ensure a healthy and constructive development of the best multimedia technologies, this workshop offers a space to discuss how to develop ethical, fair, unbiased, representative, and transparent multimedia models, bringing together researchers from different areas to present computational solutions to these issues.

Manipulating visual attributes of images through human-written text is a very challenging task. On the one hand, models have to learn the manipulation without the ground truth of the desired output. On the other hand, models have to deal with the inherent ambiguity of natural language. Previous research usually requires either the user to describe all the characteristics of the desired image or to use richly-annotated image captioning datasets. In this work, we propose a novel unsupervised approach, based on image-to-image translation, that alters the attributes of a given image through a command-like sentence such as "change the hair color to black". Contrarily to state-of-the-art approaches, our model does not require a human-annotated dataset nor a textual description of all the attributes of the desired image, but only those that have to be modified. Our proposed model disentangles the image content from the visual attributes, and it learns to modify the latter using the textual description, before generating a new image from the content and the modified attribute representation. Because text might be inherently ambiguous (blond hair may refer to different shadows of blond, e.g. golden, icy, sandy), our method generates multiple stochastic versions of the same translation. Experiments show that the proposed model achieves promising performances on two large-scale public datasets: CelebA and CUB. We believe our approach will pave the way to new avenues of research combining textual and speech commands with visual attributes.

Deep learning revolutionized data science, and recently its popularity has grown exponentially, as did the amount of papers employing deep networks. Vision tasks, such as human pose estimation, did not escape from this trend. There is a large number of deep models, where small changes in the network architecture, or in the data pre-processing, together with the stochastic nature of the optimization procedures, produce notably different results, making extremely difficult to sift methods that significantly outperform others. This situation motivates the current study, in which we perform a systematic evaluation and statistical analysis of vanilla deep regression, i.e., convolutional neural networks with a linear regression top layer. This is the first comprehensive analysis of deep regression techniques. We perform experiments on four vision problems, and report confidence intervals for the median performance as well as the statistical significance of the results, if any. Surprisingly, the variability due to different data pre-processing procedures generally eclipses the variability due to modifications in the network architecture. Our results reinforce the hypothesis according to which, in general, a general-purpose network (e.g., VGG-16 or ResNet-50) adequately tuned can yield results close to the state-of-the-art without having to resort to more complex and ad-hoc regression models.

Variational auto-encoders (VAEs) are deep generative latent variable models that can be used for learning the distribution of complex data. VAEs have been successfully used to learn a probabilistic prior over speech signals, which is then used to perform speech enhancement. One advantage of this generative approach is that it does not require pairs of clean and noisy speech signals at training. In this article, we propose audio-visual variants of VAEs for single-channel and speaker-independent speech enhancement. We develop a conditional VAE (CVAE) where the audio speech generative process is conditioned on visual information of the lip region. At test time, the audio-visual speech generative model is combined with a noise model based on nonnegative matrix factorization, and speech enhancement relies on a Monte Carlo expectation-maximization algorithm. Experiments are conducted with the recently published NTCD-TIMIT dataset as well as the GRID corpus. The results confirm that the proposed audio-visual CVAE effectively fuses audio and visual information, and it improves the speech enhancement performance compared with the audio-only VAE model, especially when the speech signal is highly corrupted by noise. We also show that the proposed unsupervised audio-visual speech enhancement approach outperforms a state-of-the-art supervised deep learning method.

While several research studies have focused on analyzing human behavior and, in particular, emotional signals from visual data, the problem of synthesizing face video sequences with specific attributes (e.g. age, facial expressions) received much less attention. This paper proposes 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. Our framework consists of two main building blocks: an image generator and a frame sequence generator. The image generator is implemented as a deep neural model which combines generative adversarial networks and variational auto-encoders, while the sequence generator is a label-conditioned recurrent neural network. In the proposed framework, given as input a neural face and a label, the sequence generator outputs a set of hidden representations with smooth transitions corresponding to video frames. Then, the image generator is used to decode the hidden representations into the actual face images. To impose that the net generates videos consistent with the given label, a novel identity adversarial loss is proposed. Our experimental results demonstrate the effectiveness of the framework and the advantage of introducing an adversarial component into recurrent models for face video generation.

Robust multi-person tracking with robots opens the door to analysing engagement and social signals in real-world environments. Multi-person scenarios are charaterised by (i) a time-varying number of people, (ii) intermittent auditory (\eg speech turns) and visual cues (\eg person appearing/disappearing) and (iii) impact of the robot actions in perception. The various sensors (cameras and microphones) available for perception, provide a rich flow of information of intermittent and complementary nature. How to jointly exploit these cues to tackle the multi-person tracking problem with an autonomous system has been an intense research line of the Perception Team in the past few years. In this demo we want to present our, now mature, achievements in the field, and demonstrate two robotic systems able to track multiple persons using auditory and visual cues, when they are available. We will bring the two robots and the necessary computing resources with us, as well as the required presentation materials to discuss the models, methods and tools supporting this technology with the attendants.

The series of FAT* events aim at bringing together researchers and practitioners interested in fairness, accountability, and transparency of computational methods. The FAT/MM workshop focuses on addressing these issues in the Multimedia field. Multimedia computing technologies operate today at an unprecedented scale, with a growing community of scientists interested in multimedia models, tools and applications. Such continued growth has great implications not only for the scientific community, but also for the society as a whole. Typical risks of large-scale computational models include model bias and algorithmic discrimination. These risks become particularly prominent in the multimedia field, which historically has been focusing on user-centered technologies. To ensure a healthy and constructive development of the best multimedia technologies, this workshop offers a space to discuss how to develop fair, unbiased, representative, and transparent multimedia models, bringing together researchers from different areas to present computational solutions to these issues.

In this letter, we address the problem of simultaneously tracking several moving audio sources, namely the problem of estimating source trajectories from a sequence of observed features. We propose to use the von Mises distribution to model audio-source directions of arrival with circular random variables. This leads to a Bayesian filtering formulation, which is intractable because of the combinatorial explosion of associating observed variables with latent variables, over time. We propose a variational approximation of the filtering distribution. We infer a variational expectation-maximization algorithm that is both computationally tractable and time efficient. We propose an audio-source birth method that favors smooth source trajectories and which is used both to initialize the number of active sources and to detect new sources. We perform experiments with the recently released LOCATA dataset comprising two moving sources and a moving microphone array mounted onto a robot.

Multimedia scientists have largely focused their research on the recognition of tangible properties of data such as objects and scenes. Recently, the field has started evolving toward the modeling of more complex properties. For example, the understanding of social, affective, and subjective attributes of visual data has attracted the attention of many research teams at the crossroads of computer vision, multimedia, and social sciences. These intangible attributes include, for example, visual beauty, video popularity, or user behavior. Multiple, diverse challenges arise when modeling such properties from multimedia data. The sections concern technical aspects such as reliable groundtruth collection, the effective learning of subjective properties, or the impact of context in subjective perception; see Refs. [2] and [3].

Recent works in computer vision and multimedia have shown that image memorability can be automatically inferred exploiting powerful deep-learning models. This article advances the state of the art in this area by addressing a novel and more challenging issue: “ Given an arbitrary input image, can we make it more memorable? ” To tackle this problem, we introduce an approach based on an editing-by-applying-filters paradigm: given an input image, we propose to automatically retrieve a set of “style seeds,” i.e., a set of style images that, applied to the input image through a neural style transfer algorithm, provide the highest increase in memorability. We show the effectiveness of the proposed approach with experiments on the publicly available LaMem dataset, performing both a quantitative evaluation and a user study. To demonstrate the flexibility of the proposed framework, we also analyze the impact of different implementation choices, such as using different state-of-the-art neural style transfer methods. Finally, we show several qualitative results to provide additional insights on the link between image style and memorability.

Each smile is unique: one person surely smiles in different ways (e.g. closing/opening the eyes or mouth). Given one input image of a neutral face, can we generate multiple smile videos with distinctive characteristics? To tackle this one-to-many video generation problem, we propose a novel deep learning architecture named Conditional Multi-Mode Network (CMM-Net). To better encode the dynamics of facial expressions, CMM-Net explicitly exploits facial landmarks for generating smile sequences. Specifically, a variational auto-encoder is used to learn a facial landmark embedding. This single embedding is then exploited by a conditional recurrent network which generates a landmark embedding sequence conditioned on a specific expression (e.g. spontaneous smile). Next, the generated landmark embeddings are fed into a multi-mode recurrent landmark generator, producing a set of landmark sequences still associated to the given smile class but clearly distinct from each other. Finally, these landmark sequences are translated into face videos. Our experimental results demonstrate the effectiveness of our CMM-Net in generating realistic videos of multiple smile expressions.

Multiple-speaker tracking is a crucial task for many applications. In real-world scenarios, exploiting the complementarity between auditory and visual data enables to track people outside the visual field of view. However, practical methods must be robust to changes in acoustic conditions, e.g. reverberation. We investigate how to combine state-of-the-art audio-source localization techniques with Bayesian multi-person tracking. Our experiments demonstrate that the performance of the proposed system is not affected by changes in the acoustic environment.

The series of events devoted to the computational Understanding of Subjective Attributes (e.g. beauty, sentiment) of Data (USAD)provide a complementary perspective to the analysis of tangible properties (objects, scenes), which overwhelmingly covered the spectra of applications in multimedia. Partly fostered by the wide-spread usage of social media, the analysis of subjective attributes has attracted lots of attention in the recent years, and many research teams at the crossroads of multimedia, computer vision and social sciences, devoted time and effort to this topic. Among the subjective attributes there are those assessed by individuals (e.g. safety,interestingness, evoked emotions [2], memorability [3]) as well as aggregated emergent properties (such as popularity or virality [1]).This edition of the workshop (see below for the workshop's history)is devoted to the multimodal recognition of evoked emotions (EE).

In this paper, we address the problem of learning robust cross-domain representations for sketch-based image retrieval (SBIR). While, most SBIR approaches focus on extracting low- and mid-level descriptors for direct feature matching, recent works have shown the benefit of learning coupled feature representations to describe data from two related sources. However, cross-domain representation learning methods are typically cast into non-convex minimization problems that are difficult to optimize, leading to unsatisfactory performance. Inspired by self-paced learning (SPL), a learning methodology designed to overcome convergence issues related to local optima by exploiting the samples in a meaningful order (i.e., easy to hard), we introduce the cross-paced partial curriculum learning (CPPCL) framework. Compared with existing SPL methods which only consider a single modality and cannot deal with prior knowledge, CPPCL is 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. In addition, thanks to the learned dictionaries, we demonstrate that the proposed CPPCL embeds robust coupled representations for SBIR. Our approach is extensively evaluated on four publicly available datasets (i.e., CUFS, Flickr15K, QueenMary SBIR, and TU-Berlin Extension datasets), showing superior performance over competing SBIR methods.

In our overly-connected world, the automatic recognition of virality - the quality of an image or video to be rapidly and widely spread in social networks - is of crucial importance, and has recently awaken the interest of the computer vision community. Concurrently, recent progress in deep learning architectures showed that global pooling strategies allow the extraction of activation maps, which highlight the parts of the image most likely to contain instances of a certain class. We extend this concept by introducing a pooling layer that learns the size of the support area to be averaged: the learned top-N average (LENA) pooling. We hypothesize that the latent concepts (feature maps) describing virality may require such a rich pooling strategy. We assess the effectiveness of the LENA layer by appending it on top of a convolutional siamese architecture and evaluate its performance on the task of predicting and localizing virality. We report experiments on two publicly available datasets annotated for virality and show that our method outperforms state-of-the-art approaches.

We present a probabilistic model for joint source separation and diarisation of multichannel convolutive speech mixtures. We build upon the framework of local Gaussian model (LGM) with non-negative matrix factorization (NMF). The diarisation is introduced as a temporal labeling of each source in the mix as active or inactive at the short-term frame level. We devise an EM algorithm in which the source separation process is aided by the diarisation state, since the latter indicates the sources actually present in the mixture. The diarisation state is tracked with a Hidden Markov Model (HMM) with emission probabilities calculated from the estimated source signals. The proposed EM has separation performance comparable with a state-of-the-art LGM NMF method, while outperforming a state-of-the-art speaker diarisation pipeline.

Multi-person tracking with a robotic platform is one of the cornerstones of human-robot interaction. Challenges arise from occlusions, appearance changes and a time-varying number of people. Furthermore, the final system is constrained by the hardware platform: low computational capacity and limited field-of-view. In this paper, we propose a novel method to simultaneously track a time-varying number of persons in three-dimensions and perform visual servoing. The complementary nature of the tracking and visual servoing enables the system to: (i) track several persons while compensating for large ego-movements and (ii) visually control the robot to keep a selected person of interest within the field of view. We propose a variational Bayesian formulation allowing us to effectively solve the inference problem through the use of closed-form solutions. Importantly, this leads to a computationally efficient procedure that runs at 10 FPS. The experiments on the NAO-MPVS dataset confirm the importance of using visual servoing for tracking multiple persons.

Recent works have shown that it is possible to automatically predict intrinsic image properties like memorability. In this paper, we take a step forward addressing the question: "Can we make an image more memorable?". Methods for automatically increasing image memorability would have an impact in many application fields like education, gaming or advertising. Our work is inspired by the popular editing-by-applying-filters paradigm adopted in photo editing applications, like Instagram and Prisma. In this context, the problem of increasing image memorability maps to that of retrieving ``memorabilizing'' filters or style ``seeds''. Still, users generally have to go through most of the available filters before finding the desired solution, thus turning the editing process into a resource and time consuming task. In this work, we show 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. Our approach leverages from recent advances in the field of image synthesis and adopts a deep architecture for generating a memorable picture from a given input image and a style seed. Importantly, to automatically select the best style a novel learning-based solution, also relying on deep models, is proposed. Our experimental evaluation, conducted on publicly available benchmarks, demonstrates the effectiveness of the proposed approach for generating memorable images through automatic style seed selection.

Multimedia scientists have largely focused their research on the recognition of tangible properties of data, such as objects and scenes. Recently, the field has started evolving towards the modeling of more complex properties. For example, the understanding of social, affective and subjective attributes of data has attracted the attention of many research teams at the crossroads of computer vision, multimedia, and social sciences. These intangible attributes include, for example, visual beauty, video popularity, or user behavior. Multiple, diverse challenges arise when modeling such properties from multimedia data. Issues concern technical aspects such as reliable groundtruth collection, the effective learning of subjective properties, or the impact of context in subjective perception. The first edition of the ACM MM'17 MUSA2 workshop has gathered together high-quality research works focusing on the computational understanding of intangible properties from multimodal data, including visual emotions, user intent, human relationships, and personality.

Recent works have shown that exploiting multi-scale representations deeply learned via convolutional neural networks (CNN) is of tremendous importance for accurate contour detection. This paper presents a novel approach for predicting contours which advances the state of the art in two fundamental aspects, i.e. multi-scale feature generation and fusion. Different from previous works directly considering multi-scale feature maps obtained from the inner layers of a primary CNN architecture, we introduce a hierarchical deep model which produces more rich and complementary representations. Furthermore, to refine and robustly fuse the representations learned at different scales, the novel Attention-Gated Conditional Random Fields (AG-CRFs) are proposed. The experiments ran on two publicly available datasets (BSDS500 and NYUDv2) demonstrate the effectiveness of the latent AG-CRF model and of the overall hierarchical framework.

This paper addresses the adaptation of an acoustic-articulatory inversion model of a reference speaker to the voice of another source speaker, using a limited amount of audio-only data. In this study, the articulatory-acoustic relationship of the reference speaker is modeled by a Gaussian mixture model and inference of articulatory data from acoustic data is made by the associated Gaussian mixture regression (GMR). To address speaker adaptation, we previously proposed a general framework called Cascaded-GMR (C-GMR) which decomposes the adaptation process into two consecutive steps: spectral conversion between source and reference speaker and acoustic-articulatory inversion of converted spectral trajectories. In particular, we proposed the integrated C-GMR technique (IC-GMR) in which both steps are tied together in the same probabilistic model. In this paper, we extend the C-GMR framework with another model called Joint-GMR (J-GMR). Contrary to the IC-GMR, this model aims at exploiting all potential acoustic-articulatory relationships, including those between the source speaker's acoustics and the reference speaker's articulation. We present the full derivation of the exact expectation-maximization (EM) training algorithm for the J-GMR. It exploits the missing data methodology of machine learning to deal with limited adaptation data. We provide an extensive evaluation of the J-GMR on both synthetic acoustic-articulatory data and on the multispeaker MOCHA EMA database. We compare the J-GMR performance to other models of the C-GMR framework, notably the IC-GMR, and discuss their respective merits.

Advanced computer vision and machine learning techniques tried to automatically categorize the emotions elicited by abstract paintings with limited success. Since the annotation of the emotional content is highly resourceconsuming, datasets of abstract paintings are either constrained in size or partially annotated. Consequently, it is natural to address the targeted task within a transductive framework. Intuitively, the use of multi-label classification techniques is desirable so to synergically exploit the relations between multiple latent variables, such as emotional content, technique, author, etc. A very popular approach for transductive multi-label recognition under linear classification settings is matrix completion. In this study we introduce non-linear matrix completion (NLMC), thus extending classical linear matrix completion techniques to the non-linear case. Together with the theory grounding the model, we propose an efficient optimization solver. As shown by our extensive experimental validation on two publicly available datasets, NLMC outperforms state-of-the-art methods when recognizing emotions from abstract paintings.

Recent studies in computer vision have shown that, while practically invisible to a human observer, skin color changes due to blood flow can be captured on face videos and, surprisingly, be used to estimate the heart rate (HR). While considerable progress has been made in the last few years, still many issues remain open. In particular, state of-the-art approaches are not robust enough to operate in natural conditions (e.g. in case of spontaneous movements, facial expressions, or illumination changes). Opposite to previous approaches that estimate the HR by processing all the skin pixels inside a fixed region of interest, we introduce a strategy to dynamically select face regions useful for robust HR estimation. Our approach, inspired by recent advances on matrix completion theory, allows us to predict the HR while simultaneously discover the best regions of the face to be used for estimation. Thorough experimental evaluation conducted on public benchmarks suggests that the proposed approach significantly outperforms state-of the-art HR estimation methods in naturalistic conditions.

In this paper we present a new statistical model for the power spectral density (PSD) of an audio signal and its application to multichannel audio source separation (MASS). The source signal is modeled with the local Gaussian model (LGM) and we propose to model its variance with an inverse-Gamma distribution, whose scale parameter is factorized as a rank-1 model. We discuss the interest of this approach and evaluate it in a MASS task with underdetermined convolutive mixtures. For this aim, we derive a variational EM algorithm for parameter estimation and source inference. The proposed model shows a benefit in source separation performance compared to a state-of-the-art LGM NMF-based technique.

Several applications benefit from learning coupled representations able to describe data from multiple sources. For instance, cross-domain dictionary learning methods demonstrated to be particularly effective. In this paper we introduce Multi-Paced Dictionary Learning (MPDL) and propose an instantiation of it under the framework of cross-domain dictionary learning. MPDL is inspired by previous works on self-paced learning, a framework able to enhance the accuracy of conventional learning models by presenting the training data in a meaningful order, i.e. easy samples are provided first. However, most of existing self-paced learning methods only consider a single modality, while MPDL is specifically designed to assess the learning pace when data from multiple sources are available. We present the model and propose an efficient algorithm to learn the dictionaries and codes. The approach is validated via experiments on two different tasks, namely cross-media retrieval and sketch-to-photo face recognition, using publicly available datasets.

While vital for handling most multimedia and computer vision problems, collecting large scale fully annotated datasets is a resource-consuming, often unaffordable task. Indeed, on the one hand datasets need to be large and variate enough so that learning strategies can successfully exploit the variability inherently present in real data, but on the other hand they should be small enough so that they can be fully annotated at a reasonable cost. With the overwhelming success of (deep) learning methods, the traditional problem of balancing between dataset dimensions and resources needed for annotations became a full-fledged dilemma. In this context, methodological approaches able to deal with partially described data sets represent a one-of-a-kind opportunity to find the right balance between data variability and resource-consumption in annotation. These include methods able to deal with noisy, weak or partial annotations. In this tutorial we will present several recent methodologies addressing different visual tasks under the assumption of noisy, weakly annotated data sets.

In the last few years, the query-by-visual-example paradigm gained popularity, specially for content based retrieval systems. As sketches represent a natural way of expressing a synthetic query, recent research efforts focused on developing algorithmic solutions to address the sketch-based image retrieval (SBIR) problem. Within this context, we propose a novel approach for SBIR that, unlike previous methods, is able to exploit the visual complexity inherently present in sketches and images. We introduce 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. We propose an instantiation of this paradigm within the framework of coupled dictionary learning to address the SBIR task. We also present an efficient algorithm to learn the dictionaries and the codes, and to pace the learning combining the reconstruction error, the prior knowledge suggested by the partial curricula and the cross-domain code coherence. In order to evaluate the proposed approach, we report an extensive experimental validation showing that the proposed method outperforms the state-of-the-art in coupled dictionary learning and in SBIR on three different publicly available datasets.

Studying free-standing conversational groups (FCGs) in unstructured social settings (e.g., cocktail party ) is gratifying due to the wealth of information available at the group (mining social networks) and individual (recognizing native behavioral and personality traits) levels. However, analyzing social scenes involving FCGs is also highly challenging due to the difficulty in extracting behavioral cues such as target locations, their speaking activity and head/body pose due to crowdedness and presence of extreme occlusions. To this end, we propose SALSA, a novel dataset facilitating multimodal and Synergetic sociAL Scene Analysis, and make two main contributions to research on automated social interaction analysis: (1) SALSA records social interactions among 18 participants in a natural, indoor environment for over 60 minutes, under the poster presentation and cocktail party contexts presenting difficulties in the form of low-resolution images, lighting variations, numerous occlusions, reverberations and interfering sound sources; (2) To alleviate these problems we facilitate multimodal analysis by recording the social interplay using four static surveillance cameras and sociometric badges worn by each participant, comprising the microphone, accelerometer, bluetooth and infrared sensors. In addition to raw data, we also provide annotations concerning individuals' personality as well as their position, head, body orientation and F-formation information over the entire event duration. Through extensive experiments with state-of-the-art approaches, we show (a) the limitations of current methods and (b) how the recorded multiple cues synergetically aid automatic analysis of social interactions. SALSA is available at http://tev.fbk.eu/salsa.

Data clustering has received a lot of attention and numerous methods, algorithms and software packages are available. Among these techniques, parametric finite-mixture models play a central role due to their interesting mathematical properties and to the existence of maximum-likelihood estimators based on expectation-maximization (EM). In this paper we propose a new mixture model that associates a weight with each observed point. We introduce the weighted-data Gaussian mixture and we derive two EM algorithms. The first one considers a fixed weight for each observation. The second one treats each weight as a random variable following a gamma distribution. We propose a model selection method based on a minimum message length criterion, provide a weight initialization strategy, and validate the proposed algorithms by comparing them with several state of the art parametric and non-parametric clustering techniques. We also demonstrate the effectiveness and robustness of the proposed clustering technique in the presence of heterogeneous data, namely audio-visual scene analysis.

This paper addresses the problem of separating audio sources from time-varying convolutive mixtures. We propose a probabilistic framework based on the local complex-Gaussian model combined with non-negative matrix factorization. The time-varying mixing filters are modeled by a continuous temporal stochastic process. We present a variational expectation-maximization (VEM) algorithm that employs a Kalman smoother to estimate the time-varying mixing matrix, and that jointly estimate the source parameters. The sound sources are then separated by Wiener filters constructed with the estimators provided by the VEM algorithm. Extensive experiments on simulated data show that the proposed method outperforms a blockwise version of a state-of-the-art baseline method.

During natural social gatherings, humans tend to organize themselves in the so-called free-standing conversational groups. In this context, robust head and body pose estimates can facilitate the higher-level description of the ongoing interplay. Importantly, visual information typically obtained with a distributed camera network might not suffice to achieve the robustness sought. In this line of thought, recent advances in wearable sensing technology open the door to multimodal and richer information flows. In this paper we propose to cast the head and body pose estimation problem into a matrix completion task. We introduce 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. We report results on the novel and challenging SALSA dataset, containing visual, auditory and infrared recordings of 18 people interacting in a regular indoor environment. We demonstrate the soundness of the proposed method and the usability for higher-level tasks such as the detection of F-formations and the discovery of social attention attractors.

This paper addresses the adaptation of an acoustic-articulatory model of a reference speaker to the voice of another speaker, using a limited amount of audio-only data. In the context of pronunciation training, a virtual talking head displaying the internal speech articulators (e.g., the tongue) could be automatically animated by means of such a model using only the speaker's voice. In this study, the articulatory-acoustic relationship of the reference speaker is modeled by a gaussian mixture model (GMM). To address the speaker adaptation problem, we propose a new framework called cascaded Gaussian mixture regression (C-GMR), and derive two implementations. The first one, referred to as Split-C-GMR, is a straightforward chaining of two distinct GMRs: one mapping the acoustic features of the source speaker into the acoustic space of the reference speaker, and the other estimating the articulatory trajectories with the reference model. In the second implementation, referred to as Integrated-C-GMR, the two mapping steps are tied together in a single probabilistic model. For this latter model, we present the full derivation of the exact EM training algorithm, that explicitly exploits the missing data methodology of machine learning. Other adaptation schemes based on maximum-a posteriori (MAP), maximum likelihood linear regression (MLLR) and direct cross-speaker acoustic-to-articulatory GMR are also investigated. Experiments conducted on two speakers for different amount of adaptation data show the interest of the proposed C-GMR techniques.

We address the problem of sound representation and classification and present results of a comparative study in the context of a domestic robotic scenario. A dataset of sounds was recorded in realistic conditions (background noise, presence of several sound sources, reverberations, etc.) using the humanoid robot NAO. An extended benchmark is carried out to test a variety of representations combined with several classifiers. We provide results obtained with the annotated dataset and we assess the methods quantitatively on the basis of their classification scores, computation times and memory requirements. The annotated dataset is publicly available at https://team.inria.fr/perception/nard/.

This paper addresses the problem of sound-source localization from time-delay estimates using arbitrarily-shaped non-coplanar microphone arrays. A novel geometric formulation is proposed, together with a thorough algebraic analysis and a global optimization solver. The proposed model is thoroughly described and evaluated. The geometric analysis, stemming from the direct acoustic propagation model, leads to necessary and sufficient conditions for a set of time delays to correspond to a unique position in the source space. Such sets of time delays are referred to as feasible sets. We formally prove that every feasible set corresponds to exactly one position in the source space, whose value can be recovered using a closed-form localization mapping. Therefore we seek for the optimal feasible set of time delays given, as input, the received microphone signals. This time delay estimation problem is naturally cast into a programming task, constrained by the feasibility conditions derived from the geometric analysis. A global branch-and-bound optimization technique is proposed to solve the problem at hand, hence estimating the best set of feasible time delays and, subsequently, localizing the sound source. Extensive experiments with both simulated and real data are reported; we compare our methodology to four state-of-the-art techniques. This comparison shows that the proposed method combined with the branch-and-bound algorithm outperforms existing methods. These in-depth geometric understanding, practical algorithms, and encouraging results, open several opportunities for future work.

The problem of choosing a classifier for audio-visual command recognition is addressed. Because such commands are culture- and user-dependant, methods need to learn new commands from a few examples. We benchmark three state-of-the-art discriminative classifiers based on bag of words and SVM. The comparison is made on monocular and monaural recordings of a publicly available dataset. We seek for the best trade off between speed, robustness and size of the training set. In the light of over 150,000 experiments, we conclude that this is a promising direction of work towards a flexible methodology that must be easily adaptable to a large variety of users.

This paper addresses the problem of audio-visual command recognition in the framework of the D-META Grand Challenge1. Temporal and non-temporal learning models are trained on visual and auditory descriptors. In order to set a proper baseline, the methods are tested on the "Robot Gestures" scenario of the publicly available RAVEL data set, following the leave-one-out cross-validation strategy. The classification-level audio-visual fusion strategy allows for compensating the errors of the unimodal (audio or vision) classifiers. The obtained results (an average audio-visual recognition rate of almost 80%) encourage us to investigate on how to further develop and improve the methodology described in this paper.

In this paper we address the problem of detecting and localizing objects that can be both seen and heard, e.g., people. This may be solved within the framework of data clustering. We propose a new multimodal clustering algorithm based on a Gaussian mixture model, where one of the modalities (visual data) is used to supervise the clustering process. This is made possible by mapping both modalities into the same metric space. To this end, we fully exploit the geometric and physical properties of an audio-visual sensor based on binocular vision and binaural hearing. We propose an EM algorithm that is theoretically well justified, intuitive, and extremely efficient from a computational point of view. This efficiency makes the method implementable on advanced platforms such as humanoid robots. We describe in detail tests and experiments performed with publicly available data sets that yield very interesting results.

In this paper we deal with the use of generalized lifting (GL) for lossy image compression. We have demonstrated in [9] the potential of the method for coding assuming complete knowledge of the pdf of the image to encode. Here, we move towards a realistic scheme that does not assume complete knowledge of the pdf. We show that a multiscale GL produces interesting results even if the pdf of the image to encode is only partially known. We target the compression of a given image class and compute an estimate of the image class pdf. This pdf is available at both encoder and decoder. A decision algorithm minimizes the overhead produced by the difference between the class pdf and the image pdf. This algorithm also removes ambiguities in the decoding process. The encoding strategy is completed using an arithmetic encoder. Results exhibit improvements over the state of the art.