Hichem Sahbi

Rémi Marsal, Florian Chabot, Angelique Loesch et al.

Self-supervised monocular depth estimation methods aim to be used in critical applications such as autonomous vehicles for environment analysis. To circumvent the potential imperfections of these approaches, a quantification of the prediction confidence is crucial to guide decision-making systems that rely on depth estimation. In this paper, we propose MonoProb, a new unsupervised monocular depth estimation method that returns an interpretable uncertainty, which means that the uncertainty reflects the expected error of the network in its depth predictions. We rethink the stereo or the structure-from-motion paradigms used to train unsupervised monocular depth models as a probabilistic problem. Within a single forward pass inference, this model provides a depth prediction and a measure of its confidence, without increasing the inference time. We then improve the performance on depth and uncertainty with a novel self-distillation loss for which a student is supervised by a pseudo ground truth that is a probability distribution on depth output by a teacher. To quantify the performance of our models we design new metrics that, unlike traditional ones, measure the absolute performance of uncertainty predictions. Our experiments highlight enhancements achieved by our method on standard depth and uncertainty metrics as well as on our tailored metrics. https://github.com/CEA-LIST/MonoProb

Hichem Sahbi

Graph convolution networks (GCNs) are currently mainstream in learning with irregular data. These models rely on message passing and attention mechanisms that capture context and node-to-node relationships. With multi-head attention, GCNs become highly accurate but oversized, and their deployment on cheap devices requires their pruning. However, pruning at high regimes usually leads to topologically inconsistent networks with weak generalization. In this paper, we devise a novel method for lightweight GCN design. Our proposed approach parses and selects subnetworks with the highest magnitudes while guaranteeing their topological consistency. The latter is obtained by selecting only accessible and co-accessible connections which actually contribute in the evaluation of the selected subnetworks. Experiments conducted on the challenging FPHA dataset show the substantial gain of our topologically consistent pruning method especially at very high pruning regimes.

Sebastien Deschamps, Hichem Sahbi

In this paper, we introduce a novel interactive satellite image change detection algorithm based on active learning. The proposed approach is iterative and asks the user (oracle) questions about the targeted changes and according to the oracle's responses updates change detections. We consider a probabilistic framework which assigns to each unlabeled sample a relevance measure modeling how critical is that sample when training change detection functions. These relevance measures are obtained by minimizing an objective function mixing diversity, representativity and uncertainty. These criteria when combined allow exploring different data modes and also refining change detections. To further explore the potential of this objective function, we consider a reinforcement learning approach that finds the best combination of diversity, representativity and uncertainty, through active learning iterations, leading to better generalization as corroborated through experiments in interactive satellite image change detection.

Hichem Sahbi

Despite the notable success of graph convolutional networks (GCNs) in skeleton-based action recognition, their performance often depends on large volumes of labeled data, which are frequently scarce in practical settings. To address this limitation, we propose a novel label-efficient GCN model. Our work makes two primary contributions. First, we develop a novel acquisition function that employs an adversarial strategy to identify a compact set of informative exemplars for labeling. This selection process balances representativeness, diversity, and uncertainty. Second, we introduce bidirectional and stable GCN architectures. These enhanced networks facilitate a more effective mapping between the ambient and latent data spaces, enabling a better understanding of the learned exemplar distribution. Extensive evaluations on two challenging skeleton-based action recognition benchmarks reveal significant improvements achieved by our label-efficient GCNs compared to prior work.

Mingyuan Jiu, Hailong Zhu, Wenchuan Wei et al.

Context modeling is crucial for visual recognition, enabling highly discriminative image representations by integrating both intrinsic and extrinsic relationships between objects and labels in images. A limitation in current approaches is their focus on basic geometric relationships or localized features, often neglecting cross-scale contextual interactions between objects. This paper introduces the Deep Panoptic Context Aggregation Network (PanCAN), a novel approach that hierarchically integrates multi-order geometric contexts through cross-scale feature aggregation in a high-dimensional Hilbert space. Specifically, PanCAN learns multi-order neighborhood relationships at each scale by combining random walks with an attention mechanism. Modules from different scales are cascaded, where salient anchors at a finer scale are selected and their neighborhood features are dynamically fused via attention. This enables effective cross-scale modeling that significantly enhances complex scene understanding by combining multi-order and cross-scale context-aware features. Extensive multi-label classification experiments on NUS-WIDE, PASCAL VOC2007, and MS-COCO benchmarks demonstrate that PanCAN consistently achieves competitive results, outperforming state-of-the-art techniques in both quantitative and qualitative evaluations, thereby substantially improving multi-label classification performance.

Hichem Sahbi, Sebastien Deschamps

Change detection in satellite imagery seeks to find occurrences of targeted changes in a given scene taken at different instants. This task has several applications ranging from land-cover mapping, to anthropogenic activity monitory as well as climate change and natural hazard damage assessment. However, change detection is highly challenging due to the acquisition conditions and also to the subjectivity of changes. In this paper, we devise a novel algorithm for change detection based on active learning. The proposed method is based on a question and answer model that probes an oracle (user) about the relevance of changes only on a small set of critical images (referred to as virtual exemplars), and according to oracle's responses updates deep neural network (DNN) classifiers. The main contribution resides in a novel adversarial model that allows learning the most representative, diverse and uncertain virtual exemplars (as inverted preimages of the trained DNNs) that challenge (the most) the trained DNNs, and this leads to a better re-estimate of these networks in the subsequent iterations of active learning. Experiments show the out-performance of our proposed deep-net inversion against the related work.

Hichem Sahbi

Magnitude pruning is one of the mainstream methods in lightweight architecture design whose goal is to extract subnetworks with the largest weight connections. This method is known to be successful, but under very high pruning regimes, it suffers from topological inconsistency which renders the extracted subnetworks disconnected, and this hinders their generalization ability. In this paper, we devise a novel magnitude pruning method that allows extracting subnetworks while guarantying their topological consistency. The latter ensures that only accessible and co-accessible -- impactful -- connections are kept in the resulting lightweight networks. Our solution is based on a novel reparametrization and two supervisory bi-directional networks which implement accessibility/co-accessibility and guarantee that only connected subnetworks will be selected during training. This solution allows enhancing generalization significantly, under very high pruning regimes, as corroborated through extensive experiments, involving graph convolutional networks, on the challenging task of skeleton-based action recognition.

Hichem Sahbi, Sebastien Deschamps

Satellite image change detection aims at finding occurrences of targeted changes in a given scene taken at different instants. This task is highly challenging due to the acquisition conditions and also to the subjectivity of changes. In this paper, we investigate satellite image change detection using active learning. Our method is interactive and relies on a question and answer model which asks the oracle (user) questions about the most informative display (dubbed as virtual exemplars), and according to the user's responses, updates change detections. The main contribution of our method consists in a novel adversarial model that allows frugally probing the oracle with only the most representative, diverse and uncertain virtual exemplars. The latter are learned to challenge the most the trained change decision criteria which ultimately leads to a better re-estimate of these criteria in the following iterations of active learning. Conducted experiments show the out-performance of our proposed adversarial display model against other display strategies as well as the related work.

Hichem Sahbi

In this paper, we design lightweight graph convolutional networks (GCNs) using a particular class of regularizers, dubbed as phase-field models (PFMs). PFMs exhibit a bi-phase behavior using a particular ultra-local term that allows training both the topology and the weight parameters of GCNs as a part of a single "end-to-end" optimization problem. Our proposed solution also relies on a reparametrization that pushes the mask of the topology towards binary values leading to effective topology selection and high generalization while implementing any targeted pruning rate. Both masks and weights share the same set of latent variables and this further enhances the generalization power of the resulting lightweight GCNs. Extensive experiments conducted on the challenging task of skeleton-based recognition show the outperformance of PFMs against other staple regularizers as well as related lightweight design methods.

Sebastien Deschamps, Hichem Sahbi

Most of the existing learning models, particularly deep neural networks, are reliant on large datasets whose hand-labeling is expensive and time demanding. A current trend is to make the learning of these models frugal and less dependent on large collections of labeled data. Among the existing solutions, deep active learning is currently witnessing a major interest and its purpose is to train deep networks using as few labeled samples as possible. However, the success of active learning is highly dependent on how critical are these samples when training models. In this paper, we devise a novel active learning approach for label-efficient training. The proposed method is iterative and aims at minimizing a constrained objective function that mixes diversity, representativity and uncertainty criteria. The proposed approach is probabilistic and unifies all these criteria in a single objective function whose solution models the probability of relevance of samples (i.e., how critical) when learning a decision function. We also introduce a novel weighting mechanism based on reinforcement learning, which adaptively balances these criteria at each training iteration, using a particular stateless Q-learning model. Extensive experiments conducted on staple image classification data, including Object-DOTA, show the effectiveness of our proposed model w.r.t. several baselines including random, uncertainty and flat as well as other work.

Hichem Sahbi, Sebastien Deschamps

In this paper, we devise a novel interactive satellite image change detection algorithm based on active learning. The proposed framework is iterative and relies on a question and answer model which asks the oracle (user) questions about the most informative display (subset of critical images), and according to the user's responses, updates change detections. The contribution of our framework resides in a novel display model which selects the most representative and diverse virtual exemplars that adversely challenge the learned change detection functions, thereby leading to highly discriminating functions in the subsequent iterations of active learning. Extensive experiments, conducted on the challenging task of interactive satellite image change detection, show the superiority of the proposed virtual display model against the related work.

Jie Mei, Mingyuan Jiu, Hichem Sahbi et al.

Few-shot detection is a major task in pattern recognition which seeks to localize objects using models trained with few labeled data. One of the mainstream few-shot methods is transfer learning which consists in pretraining a detection model in a source domain prior to its fine-tuning in a target domain. However, it is challenging for fine-tuned models to effectively identify new classes in the target domain, particularly when the underlying labeled training data are scarce. In this paper, we devise a novel sparse context transformer (SCT) that effectively leverages object knowledge in the source domain, and automatically learns a sparse context from only few training images in the target domain. As a result, it combines different relevant clues in order to enhance the discrimination power of the learned detectors and reduce class confusion. We evaluate the proposed method on two challenging few-shot object detection benchmarks, and empirical results show that the proposed method obtains competitive performance compared to the related state-of-the-art.

Victor Enescu, Hichem Sahbi

Continual or incremental learning holds tremendous potential in deep learning with different challenges including catastrophic forgetting. The advent of powerful foundation and generative models has propelled this paradigm even further, making it one of the most viable solution to train these models. However, one of the persisting issues lies in the increasing volume of data particularly with replay-based methods. This growth introduces challenges with scalability since continuously expanding data becomes increasingly demanding as the number of tasks grows. In this paper, we attenuate this issue by devising a novel replay-free incremental learning model based on Variational Autoencoders (VAEs). The main contribution of this work includes (i) a novel incremental generative modelling, built upon a well designed multi-modal latent space, and also (ii) an orthogonality criterion that mitigates catastrophic forgetting of the learned VAEs. The proposed method considers two variants of these VAEs: static and dynamic with no (or at most a controlled) growth in the number of parameters. Extensive experiments show that our method is (at least) an order of magnitude more ``memory-frugal'' compared to the closely related works while achieving SOTA accuracy scores.

Mingyuan Jiu, Xuguang Song, Hichem Sahbi et al.

Deep learning is currently reaching outstanding performances on different tasks, including image classification, especially when using large neural networks. The success of these models is tributary to the availability of large collections of labeled training data. In many real-world scenarios, labeled data are scarce, and their hand-labeling is time, effort and cost demanding. Active learning is an alternative paradigm that mitigates the effort in hand-labeling data, where only a small fraction is iteratively selected from a large pool of unlabeled data, and annotated by an expert (a.k.a oracle), and eventually used to update the learning models. However, existing active learning solutions are dependent on handcrafted strategies that may fail in highly variable learning environments (datasets, scenarios, etc). In this work, we devise an adaptive active learning method based on Markov Decision Process (MDP). Our framework leverages deep reinforcement learning and active learning together with a Deep Deterministic Policy Gradient (DDPG) in order to dynamically adapt sample selection strategies to the oracle's feedback and the learning environment. Extensive experiments conducted on three different image classification benchmarks show superior performances against several existing active learning strategies.

Hichem Sahbi

Skeleton-based action recognition is a hotspot in image processing. A key challenge of this task lies in its dependence on large, manually labeled datasets whose acquisition is costly and time-consuming. This paper devises a novel, label-efficient method for skeleton-based action recognition using graph convolutional networks (GCNs). The contribution of the proposed method resides in learning a novel acquisition function -- scoring the most informative subsets for labeling -- as the optimum of an objective function mixing data representativity, diversity and uncertainty. We also extend this approach by learning the most informative subsets using an invertible GCN which allows mapping data from ambient to latent spaces where the inherent distribution of the data is more easily captured. Extensive experiments, conducted on two challenging skeleton-based recognition datasets, show the effectiveness and the outperformance of our label-frugal GCNs against the related work.

Hichem Sahbi

This paper devises a novel interactive satellite image change detection algorithm based on active learning. Our framework employs an iterative process that leverages a question-and-answer model. This model queries the oracle (user) about the labels of a small subset of images (dubbed as display), and based on the oracle's responses, change detection model is dynamically updated. The main contribution of our framework resides in a novel invertible network that allows augmenting displays, by mapping them from highly nonlinear input spaces to latent ones, where augmentation transformations become linear and more tractable. The resulting augmented data are afterwards mapped back to the input space, and used to retrain more effective change detection criteria in the subsequent iterations of active learning. Experimental results demonstrate superior performance of our proposed method compared to the related work.

Hichem Sahbi

Change detection is a major task in remote sensing which consists in finding all the occurrences of changes in multi-temporal satellite or aerial images. The success of existing methods, and particularly deep learning ones, is tributary to the availability of hand-labeled training data that capture the acquisition conditions and the subjectivity of the user (oracle). In this paper, we devise a novel change detection algorithm, based on active learning. The main contribution of our work resides in a new model that measures how important is each unlabeled sample, and provides an oracle with only the most critical samples (also referred to as virtual exemplars) for further labeling. These exemplars are generated, using an invertible graph convnet, as the optimum of an adversarial loss that (i) measures representativity, diversity and ambiguity of the data, and thereby (ii) challenges (the most) the current change detection criteria, leading to a better re-estimate of these criteria in the subsequent iterations of active learning. Extensive experiments show the positive impact of our label-efficient learning model against comparative methods.

Victor Enescu, Hichem Sahbi

Continual adaptation of deep generative models holds tremendous potential and critical importance, given their rapid and expanding usage in text and vision based applications. Incremental training, however, remains highly challenging due to catastrophic forgetting phenomenon, which makes it difficult for neural networks to effectively incorporate new knowledge. A common strategy consists in retraining the generative model on its own synthetic data in order to mitigate forgetting. Yet, such an approach faces two major limitations: (i) the continually increasing training time eventually becomes intractable, and (ii) reliance on synthetic data inevitably leads to long-term performance degradation, since synthetic samples lack the richness of real training data. In this paper, we attenuate these issues by designing a novel and more expressive conditioning mechanism for generative models based on low rank adaptation (LoRA), that exclusively employs rank 1 matrices, whose reparametrized matrix rank is functionally increased using carefully selected functions -- and dubbed functional LoRA: FunLoRA. Using this dynamic conditioning, the generative model is guaranteed to avoid catastrophic forgetting and needs only to be trained on data from the current task. Extensive experiments using flow-matching based models trained from scratch, showcase that our proposed parameter-efficient fine-tuning (PEFT) method surpasses prior state-of-the-art results based on diffusion models, reaching higher classification accuracy scores, while only requiring a fraction of the memory cost and sampling time.

Mingyuan Jiu, Hailong Zhu, Hichem Sahbi

Multi-label classification is a challenging task in pattern recognition. Many deep learning methods have been proposed and largely enhanced classification performance. However, most of the existing sophisticated methods ignore context in the models' learning process. Since context may provide additional cues to the learned models, it may significantly boost classification performances. In this work, we make full use of context information (namely geometrical structure of images) in order to learn better context-aware similarities (a.k.a. kernels) between images. We reformulate context-aware kernel design as a feed-forward network that outputs explicit kernel mapping features. Our obtained context-aware kernel network further leverages multiple orders of patch neighbors within different distances, resulting into a more discriminating Deep Multi-order Context-aware Kernel Network (DMCKN) for multi-label classification. We evaluate the proposed method on the challenging Corel5K and NUS-WIDE benchmarks, and empirical results show that our method obtains competitive performances against the related state-of-the-art, and both quantitative and qualitative performances corroborate its effectiveness and superiority for multi-label image classification.

Hichem Sahbi

Magnitude Pruning is a staple lightweight network design method which seeks to remove connections with the smallest magnitude. This process is either achieved in a structured or unstructured manner. While structured pruning allows reaching high efficiency, unstructured one is more flexible and leads to better accuracy, but this is achieved at the expense of low computational performance. In this paper, we devise a novel coarse-to-fine (CTF) method that gathers the advantages of structured and unstructured pruning while discarding their inconveniences to some extent. Our method relies on a novel CTF parametrization that models the mask of each connection as the Hadamard product involving four parametrizations which capture channel-wise, column-wise, row-wise and entry-wise pruning respectively. Hence, fine-grained pruning is enabled only when the coarse-grained one is disabled, and this leads to highly efficient networks while being effective. Extensive experiments conducted on the challenging task of skeleton-based recognition, using the standard SBU and FPHA datasets, show the clear advantage of our CTF approach against different baselines as well as the related work.

Hichem Sahbi

Deep neural networks (DNNs) are nowadays witnessing a major success in solving many pattern recognition tasks including skeleton-based classification. The deployment of DNNs on edge-devices, endowed with limited time and memory resources, requires designing lightweight and efficient variants of these networks. Pruning is one of the lightweight network design techniques that operate by removing unnecessary network parts, in a structured or an unstructured manner, including individual weights, neurons or even entire channels. Nonetheless, structured and unstructured pruning methods, when applied separately, may either be inefficient or ineffective. In this paper, we devise a novel semi-structured method that discards the downsides of structured and unstructured pruning while gathering their upsides to some extent. The proposed solution is based on a differentiable cascaded parametrization which combines (i) a band-stop mechanism that prunes weights depending on their magnitudes, (ii) a weight-sharing parametrization that prunes connections either individually or group-wise, and (iii) a gating mechanism which arbitrates between different group-wise and entry-wise pruning. All these cascaded parametrizations are built upon a common latent tensor which is trained end-to-end by minimizing a classification loss and a surrogate tensor rank regularizer. Extensive experiments, conducted on the challenging tasks of action and hand-gesture recognition, show the clear advantage of our proposed semi-structured pruning approach against both structured and unstructured pruning, when taken separately, as well as the related work.

Hichem Sahbi

In this paper, we devise a novel lightweight Graph Convolutional Network (GCN) design dubbed as Multi-Rate Magnitude Pruning (MRMP) that jointly trains network topology and weights. Our method is variational and proceeds by aligning the weight distribution of the learned networks with an a priori distribution. In the one hand, this allows implementing any fixed pruning rate, and also enhancing the generalization performances of the designed lightweight GCNs. In the other hand, MRMP achieves a joint training of multiple GCNs, on top of shared weights, in order to extrapolate accurate networks at any targeted pruning rate without retraining their weights. Extensive experiments conducted on the challenging task of skeleton-based recognition show a substantial gain of our lightweight GCNs particularly at very high pruning regimes.

Hichem Sahbi

We introduce a novel interactive satellite image change detection algorithm based on active learning. The proposed method is iterative and consists in frugally probing the user (oracle) about the labels of the most critical images, and according to the oracle's annotations, it updates change detection results. First, we consider a probabilistic framework which assigns to each unlabeled sample a relevance measure modeling how critical is that sample when training change detection functions. We obtain these relevance measures by minimizing an objective function mixing diversity, representativity and uncertainty. These criteria when combined allow exploring different data modes and also refining change detections. Then, we further explore the potential of this objective function, by considering a reinforcement learning approach that finds the best combination of diversity, representativity and uncertainty as well as display-sizes through active learning iterations, leading to better generalization as shown through experiments in interactive satellite image change detection.

Hichem Sahbi

Graph convolutional networks (GCNs) are nowadays becoming mainstream in solving many image processing tasks including skeleton-based recognition. Their general recipe consists in learning convolutional and attention layers that maximize classification performances. With multi-head attention, GCNs are highly accurate but oversized, and their deployment on edge devices requires their pruning. Among existing methods, magnitude pruning (MP) is relatively effective but its design is clearly suboptimal as network topology selection and weight retraining are achieved independently. In this paper, we devise a novel lightweight GCN design dubbed as Probabilistic Magnitude Pruning (PMP) that jointly trains network topology and weights. Our method is variational and proceeds by aligning the weight distribution of the learned networks with an a priori distribution. This allows implementing any fixed pruning rate, and also enhancing the generalization performances of the designed lightweight GCNs. Extensive experiments conducted on the challenging task of skeleton-based recognition show a substantial gain of our lightweight GCNs particularly at very high pruning regimes.

Robin Dupont, Mohammed Amine Alaoui, Hichem Sahbi et al.

Large and performant neural networks are often overparameterized and can be drastically reduced in size and complexity thanks to pruning. Pruning is a group of methods, which seeks to remove redundant or unnecessary weights or groups of weights in a network. These techniques allow the creation of lightweight networks, which are particularly critical in embedded or mobile applications. In this paper, we devise an alternative pruning method that allows extracting effective subnetworks from larger untrained ones. Our method is stochastic and extracts subnetworks by exploring different topologies which are sampled using Gumbel Softmax. The latter is also used to train probability distributions which measure the relevance of weights in the sampled topologies. The resulting subnetworks are further enhanced using a highly efficient rescaling mechanism that reduces training time and improves performance. Extensive experiments conducted on CIFAR show the outperformance of our subnetwork extraction method against the related work.

Hichem Sahbi

Learning graph convolutional networks (GCNs) is an emerging field which aims at generalizing convolutional operations to arbitrary non-regular domains. In particular, GCNs operating on spatial domains show superior performances compared to spectral ones, however their success is highly dependent on how the topology of input graphs is defined. In this paper, we introduce a novel framework for graph convolutional networks that learns the topological properties of graphs. The design principle of our method is based on the optimization of a constrained objective function which learns not only the usual convolutional parameters in GCNs but also a transformation basis that conveys the most relevant topological relationships in these graphs. Experiments conducted on the challenging task of skeleton-based action recognition shows the superiority of the proposed method compared to handcrafted graph design as well as the related work.

Hichem Sahbi, Haoming Zhan

Deep neural networks (DNNs) have recently achieved a great success in computer vision and several related fields. Despite such progress, current neural architectures still suffer from catastrophic interference (a.k.a. forgetting) which obstructs DNNs to learn continually. While several state-of-the-art methods have been proposed to mitigate forgetting, these existing solutions are either highly rigid (as regularization) or time/memory demanding (as replay). An intermediate class of methods, based on dynamic networks, has been proposed in the literature and provides a reasonable balance between task memorization and computational footprint. In this paper, we devise a dynamic network architecture for continual learning based on a novel forgetting-free neural block (FFNB). Training FFNB features on new tasks is achieved using a novel procedure that constrains the underlying parameters in the null-space of the previous tasks, while training classifier parameters equates to Fisher discriminant analysis. The latter provides an effective incremental process which is also optimal from a Bayesian perspective. The trained features and classifiers are further enhanced using an incremental "end-to-end" fine-tuning. Extensive experiments, conducted on different challenging classification problems, show the high effectiveness of the proposed method.

Hichem Sahbi, Sebastien Deschamps, Andrei Stoian

We introduce in this paper a novel active learning algorithm for satellite image change detection. The proposed solution is interactive and based on a question and answer model, which asks an oracle (annotator) the most informative questions about the relevance of sampled satellite image pairs, and according to the oracle's responses, updates a decision function iteratively. We investigate a novel framework which models the probability that samples are relevant; this probability is obtained by minimizing an objective function capturing representativity, diversity and ambiguity. Only data with a high probability according to these criteria are selected and displayed to the oracle for further annotation. Extensive experiments on the task of satellite image change detection after natural hazards (namely tornadoes) show the relevance of the proposed method against the related work.

Robin Dupont, Hichem Sahbi, Guillaume Michel

Pruning seeks to design lightweight architectures by removing redundant weights in overparameterized networks. Most of the existing techniques first remove structured sub-networks (filters, channels,...) and then fine-tune the resulting networks to maintain a high accuracy. However, removing a whole structure is a strong topological prior and recovering the accuracy, with fine-tuning, is highly cumbersome. In this paper, we introduce an "end-to-end" lightweight network design that achieves training and pruning simultaneously without fine-tuning. The design principle of our method relies on reparametrization that learns not only the weights but also the topological structure of the lightweight sub-network. This reparametrization acts as a prior (or regularizer) that defines pruning masks implicitly from the weights of the underlying network, without increasing the number of training parameters. Sparsity is induced with a budget loss that provides an accurate pruning. Extensive experiments conducted on the CIFAR10 and the TinyImageNet datasets, using standard architectures (namely Conv4, VGG19 and ResNet18), show compelling results without fine-tuning.

Hichem Sahbi

Spectral graph convolutional networks (GCNs) are particular deep models which aim at extending neural networks to arbitrary irregular domains. The principle of these networks consists in projecting graph signals using the eigen-decomposition of their Laplacians, then achieving filtering in the spectral domain prior to back-project the resulting filtered signals onto the input graph domain. However, the success of these operations is highly dependent on the relevance of the used Laplacians which are mostly handcrafted and this makes GCNs clearly sub-optimal. In this paper, we introduce a novel spectral GCN that learns not only the usual convolutional parameters but also the Laplacian operators. The latter are designed "end-to-end" as a part of a recursive Chebyshev decomposition with the particularity of conveying both the differential and the non-differential properties of the learned representations -- with increasing order and discrimination power -- without overparametrizing the trained GCNs. Extensive experiments, conducted on the challenging task of skeleton-based action recognition, show the generalization ability and the outperformance of our proposed Laplacian design w.r.t. different baselines (built upon handcrafted and other learned Laplacians) as well as the related work.

Hichem Sahbi

Graph convolutional networks (GCNs) aim at extending deep learning to arbitrary irregular domains, namely graphs. Their success is highly dependent on how the topology of input graphs is defined and most of the existing GCN architectures rely on predefined or handcrafted graph structures. In this paper, we introduce a novel method that learns the topology (or connectivity) of input graphs as a part of GCN design. The main contribution of our method resides in building an orthogonal connectivity basis that optimally aggregates nodes, through their neighborhood, prior to achieve convolution. Our method also considers a stochasticity criterion which acts as a regularizer that makes the learned basis and the underlying GCNs lightweight while still being highly effective. Experiments conducted on the challenging task of skeleton-based hand-gesture recognition show the high effectiveness of the learned GCNs w.r.t. the related work.

Hichem Sahbi

Learning graph convolutional networks (GCNs) is an emerging field which aims at generalizing deep learning to arbitrary non-regular domains. Most of the existing GCNs follow a neighborhood aggregation scheme, where the representation of a node is recursively obtained by aggregating its neighboring node representations using averaging or sorting operations. However, these operations are either ill-posed or weak to be discriminant or increase the number of training parameters and thereby the computational complexity and the risk of overfitting. In this paper, we introduce a novel GCN framework that achieves spatial graph convolution in a reproducing kernel Hilbert space (RKHS). The latter makes it possible to design, via implicit kernel representations, convolutional graph filters in a high dimensional and more discriminating space without increasing the number of training parameters. The particularity of our GCN model also resides in its ability to achieve convolutions without explicitly realigning nodes in the receptive fields of the learned graph filters with those of the input graphs, thereby making convolutions permutation agnostic and well defined. Experiments conducted on the challenging task of skeleton-based action recognition show the superiority of the proposed method against different baselines as well as the related work.

Mingyuan Jiu, Hichem Sahbi

Context modeling is one of the most fertile subfields of visual recognition which aims at designing discriminant image representations while incorporating their intrinsic and extrinsic relationships. However, the potential of context modeling is currently underexplored and most of the existing solutions are either context-free or restricted to simple handcrafted geometric relationships. We introduce in this paper DHCN: a novel Deep Hierarchical Context Network that leverages different sources of contexts including geometric and semantic relationships. The proposed method is based on the minimization of an objective function mixing a fidelity term, a context criterion and a regularizer. The solution of this objective function defines the architecture of a bi-level hierarchical context network; the first level of this network captures scene geometry while the second one corresponds to semantic relationships. We solve this representation learning problem by training its underlying deep network whose parameters correspond to the most influencing bi-level contextual relationships and we evaluate its performances on image annotation using the challenging ImageCLEF benchmark.

Mingyuan Jiu, Hichem Sahbi

Deep kernel map networks have shown excellent performances in various classification problems including image annotation. Their general recipe consists in aggregating several layers of singular value decompositions (SVDs) -- that map data from input spaces into high dimensional spaces -- while preserving the similarity of the underlying kernels. However, the potential of these deep map networks has not been fully explored as the original setting of these networks focuses mainly on the approximation quality of their kernels and ignores their discrimination power. In this paper, we introduce a novel "end-to-end" design for deep kernel map learning that balances the approximation quality of kernels and their discrimination power. Our method proceeds in two steps; first, layerwise SVD is applied in order to build initial deep kernel map approximations and then an "end-to-end" supervised learning is employed to further enhance their discrimination power while maintaining their efficiency. Extensive experiments, conducted on the challenging ImageCLEF annotation benchmark, show the high efficiency and the out-performance of this two-step process with respect to different related methods.

Ahmed Mazari, Hichem Sahbi

Most of the current action recognition algorithms are based on deep networks which stack multiple convolutional, pooling and fully connected layers. While convolutional and fully connected operations have been widely studied in the literature, the design of pooling operations that handle action recognition, with different sources of temporal granularity in action categories, has comparatively received less attention, and existing solutions rely mainly on max or averaging operations. The latter are clearly powerless to fully exhibit the actual temporal granularity of action categories and thereby constitute a bottleneck in classification performances. In this paper, we introduce a novel hierarchical pooling design that captures different levels of temporal granularity in action recognition. Our design principle is coarse-to-fine and achieved using a tree-structured network; as we traverse this network top-down, pooling operations are getting less invariant but timely more resolute and well localized. Learning the combination of operations in this network -- which best fits a given ground-truth -- is obtained by solving a constrained minimization problem whose solution corresponds to the distribution of weights that capture the contribution of each level (and thereby temporal granularity) in the global hierarchical pooling process. Besides being principled and well grounded, the proposed hierarchical pooling is also video-length and resolution agnostic. Extensive experiments conducted on the challenging UCF-101, HMDB-51 and JHMDB-21 databases corroborate all these statements.

Ahmed Mazari, Hichem Sahbi

In this paper, we introduce a novel hierarchical aggregation design that captures different levels of temporal granularity in action recognition. Our design principle is coarse-to-fine and achieved using a tree-structured network; as we traverse this network top-down, pooling operations are getting less invariant but timely more resolute and well localized. Learning the combination of operations in this network -- which best fits a given ground-truth -- is obtained by solving a constrained minimization problem whose solution corresponds to the distribution of weights that capture the contribution of each level (and thereby temporal granularity) in the global hierarchical pooling process. Besides being principled and well grounded, the proposed hierarchical pooling is also video-length agnostic and resilient to misalignments in actions. Extensive experiments conducted on the challenging UCF-101 database corroborate these statements.

Mingyuan Jiu, Hichem Sahbi

Context plays a crucial role in visual recognition as it provides complementary clues for different learning tasks including image classification and annotation. As the performances of these tasks are currently reaching a plateau, any extra knowledge, including context, should be leveraged in order to seek significant leaps in these performances. In the particular scenario of kernel machines, context-aware kernel design aims at learning positive semi-definite similarity functions which return high values not only when data share similar contents, but also similar structures (a.k.a contexts). However, the use of context in kernel design has not been fully explored; indeed, context in these solutions is handcrafted instead of being learned. In this paper, we introduce a novel deep network architecture that learns context in kernel design. This architecture is fully determined by the solution of an objective function mixing a content term that captures the intrinsic similarity between data, a context criterion which models their structure and a regularization term that helps designing smooth kernel network representations. The solution of this objective function defines a particular deep network architecture whose parameters correspond to different variants of learned contexts including layerwise, stationary and classwise; larger values of these parameters correspond to the most influencing contextual relationships between data. Extensive experiments conducted on the challenging ImageCLEF Photo Annotation and Corel5k benchmarks show that our deep context networks are highly effective for image classification and the learned contexts further enhance the performance of image annotation.

Hichem Sahbi

Support vector machines (SVMs) have been successful in solving many computer vision tasks including image and video category recognition especially for small and mid-scale training problems. The principle of these non-parametric models is to learn hyperplanes that separate data belonging to different classes while maximizing their margins. However, SVMs constrain the learned hyperplanes to lie in the span of support vectors, fixed/taken from training data, and this reduces their representational power and may lead to limited generalization performances. In this paper, we relax this constraint and allow the support vectors to be learned (instead of being fixed/taken from training data) in order to better fit a given classification task. Our approach, referred to as deep total variation support vector machines, is parametric and relies on a novel deep architecture that learns not only the SVM and the kernel parameters but also the support vectors, resulting into highly effective classifiers. We also show (under a particular setting of the activation functions in this deep architecture) that a large class of kernels and their combinations can be learned. Experiments conducted on the challenging task of skeleton-based action recognition show the outperformance of our deep total variation SVMs w.r.t different baselines as well as the related work.

Ahmed Mazari, Hichem Sahbi

Convolutional neural networks are nowadays witnessing a major success in different pattern recognition problems. These learning models were basically designed to handle vectorial data such as images but their extension to non-vectorial and semi-structured data (namely graphs with variable sizes, topology, etc.) remains a major challenge, though a few interesting solutions are currently emerging. In this paper, we introduce MLGCN; a novel spectral Multi-Laplacian Graph Convolutional Network. The main contribution of this method resides in a new design principle that learns graph-laplacians as convex combinations of other elementary laplacians each one dedicated to a particular topology of the input graphs. We also introduce a novel pooling operator, on graphs, that proceeds in two steps: context-dependent node expansion is achieved, followed by a global average pooling; the strength of this two-step process resides in its ability to preserve the discrimination power of nodes while achieving permutation invariance. Experiments conducted on SBU and UCF-101 datasets, show the validity of our method for the challenging task of action recognition.

Ahmed Mazari, Hichem Sahbi

Deep convolutional neural networks (CNNs) are nowadays achieving significant leaps in different pattern recognition tasks including action recognition. Current CNNs are increasingly deeper, data-hungrier and this makes their success tributary of the abundance of labeled training data. CNNs also rely on max/average pooling which reduces dimensionality of output layers and hence attenuates their sensitivity to the availability of labeled data. However, this process may dilute the information of upstream convolutional layers and thereby affect the discrimination power of the trained representations, especially when the learned categories are fine-grained. In this paper, we introduce a novel hierarchical aggregation design, for final pooling, that controls granularity of the learned representations w.r.t the actual granularity of action categories. Our solution is based on a tree-structured temporal pyramid that aggregates outputs of CNNs at different levels. Top levels of this hierarchy are dedicated to coarse categories while deep levels are more suitable to fine-grained ones. The design of our temporal pyramid is based on solving a constrained minimization problem whose solution corresponds to the distribution of weights of different representations in the temporal pyramid. Experiments conducted using the challenging UCF101 database show the relevance of our hierarchical design w.r.t other related methods.

Hichem Sahbi

We introduce in this work a novel stochastic inference process, for scene annotation and object class segmentation, based on finite state machines (FSMs). The design principle of our framework is generative and based on building, for a given scene, finite state machines that encode annotation lattices, and inference consists in finding and scoring the best configurations in these lattices. Different novel operations are defined using our FSM framework including reordering, segmentation, visual transduction, and label dependency modeling. All these operations are combined together in order to achieve annotation as well as object class segmentation.

Hichem Sahbi

Canonical correlation analysis (CCA) is a statistical learning method that seeks to build view-independent latent representations from multi-view data. This method has been successfully applied to several pattern analysis tasks such as image-to-text mapping and view-invariant object/action recognition. However, this success is highly dependent on the quality of data pairing (i.e., alignments) and mispairing adversely affects the generalization ability of the learned CCA representations. In this paper, we address the issue of alignment errors using a new variant of canonical correlation analysis referred to as alignment-agnostic (AA) CCA. Starting from erroneously paired data taken from different views, this CCA finds transformation matrices by optimizing a constrained maximization problem that mixes a data correlation term with context regularization; the particular design of these two terms mitigates the effect of alignment errors when learning the CCA transformations. Experiments conducted on multi-view tasks, including multi-temporal satellite image change detection, show that our AA CCA method is highly effective and resilient to mispairing errors.

Hichem Sahbi

Deep networks are nowadays becoming popular in many computer vision and pattern recognition tasks. Among these networks, deep kernels are particularly interesting and effective, however, their computational complexity is a major issue especially on cheap hardware resources. In this paper, we address the issue of efficient computation in deep kernel networks. We propose a novel framework that reduces dramatically the complexity of evaluating these deep kernels. Our method is based on a coarse-to-fine cascade of networks designed for efficient computation; early stages of the cascade are cheap and reject many patterns efficiently while deep stages are more expensive and accurate. The design principle of these reduced complexity networks is based on a variant of the cross-entropy criterion that reduces the complexity of the networks in the cascade while preserving all the positive responses of the original kernel network. Experiments conducted - on the challenging and time demanding change detection task, on very large satellite images - show that our proposed coarse-to-fine approach is effective and highly efficient.

Mingyuan Jiu, Hichem Sahbi

Deep kernel learning aims at designing nonlinear combinations of multiple standard elementary kernels by training deep networks. This scheme has proven to be effective, but intractable when handling large-scale datasets especially when the depth of the trained networks increases; indeed, the complexity of evaluating these networks scales quadratically w.r.t. the size of training data and linearly w.r.t. the depth of the trained networks. In this paper, we address the issue of efficient computation in Deep Kernel Networks (DKNs) by designing effective maps in the underlying Reproducing Kernel Hilbert Spaces. Given a pretrained DKN, our method builds its associated Deep Map Network (DMN) whose inner product approximates the original network while being far more efficient. The design principle of our method is greedy and achieved layer-wise, by finding maps that approximate DKNs at different (input, intermediate and output) layers. This design also considers an extra fine-tuning step based on unsupervised learning, that further enhances the generalization ability of the trained DMNs. When plugged into SVMs, these DMNs turn out to be as accurate as the underlying DKNs while being at least an order of magnitude faster on large-scale datasets, as shown through extensive experiments on the challenging ImageCLEF and COREL5k benchmarks.

Mingyuan Jiu, Hichem Sahbi

Context plays an important role in visual pattern recognition as it provides complementary clues for different learning tasks including image classification and annotation. In the particular scenario of kernel learning, the general recipe of context-based kernel design consists in learning positive semi-definite similarity functions that return high values not only when data share similar content but also similar context. However, in spite of having a positive impact on performance, the use of context in these kernel design methods has not been fully explored; indeed, context has been handcrafted instead of being learned. In this paper, we introduce a novel context-aware kernel design framework based on deep learning. Our method discriminatively learns spatial geometric context as the weights of a deep network (DN). The architecture of this network is fully determined by the solution of an objective function that mixes content, context and regularization, while the parameters of this network determine the most relevant (discriminant) parts of the learned context. We apply this context and kernel learning framework to image classification using the challenging ImageCLEF Photo Annotation benchmark; the latter shows that our deep context learning provides highly effective kernels for image classification as corroborated through extensive experiments.

Anjan Dutta, Hichem Sahbi

Graph-based methods are known to be successful in many machine learning and pattern classification tasks. These methods consider semi-structured data as graphs where nodes correspond to primitives (parts, interest points, segments, etc.) and edges characterize the relationships between these primitives. However, these non-vectorial graph data cannot be straightforwardly plugged into off-the-shelf machine learning algorithms without a preliminary step of -- explicit/implicit -- graph vectorization and embedding. This embedding process should be resilient to intra-class graph variations while being highly discriminant. In this paper, we propose a novel high-order stochastic graphlet embedding (SGE) that maps graphs into vector spaces. Our main contribution includes a new stochastic search procedure that efficiently parses a given graph and extracts/samples unlimitedly high-order graphlets. We consider these graphlets, with increasing orders, to model local primitives as well as their increasingly complex interactions. In order to build our graph representation, we measure the distribution of these graphlets into a given graph, using particular hash functions that efficiently assign sampled graphlets into isomorphic sets with a very low probability of collision. When combined with maximum margin classifiers, these graphlet-based representations have positive impact on the performance of pattern comparison and recognition as corroborated through extensive experiments using standard benchmark databases.

Anjan Dutta, Hichem Sahbi

Graph-based methods are known to be successful in many machine learning and pattern classification tasks. These methods consider semi-structured data as graphs where nodes correspond to primitives (parts, interest points, segments, etc.) and edges characterize the relationships between these primitives. However, these non-vectorial graph data cannot be straightforwardly plugged into off-the-shelf machine learning algorithms without a preliminary step of -- explicit/implicit -- graph vectorization and embedding. This embedding process should be resilient to intra-class graph variations while being highly discriminant. In this paper, we propose a novel high-order stochastic graphlet embedding (SGE) that maps graphs into vector spaces. Our main contribution includes a new stochastic search procedure that efficiently parses a given graph and extracts/samples unlimitedly high-order graphlets. We consider these graphlets, with increasing orders, to model local primitives as well as their increasingly complex interactions. In order to build our graph representation, we measure the distribution of these graphlets into a given graph, using particular hash functions that efficiently assign sampled graphlets into isomorphic sets with a very low probability of collision. When combined with maximum margin classifiers, these graphlet-based representations have positive impact on the performance of pattern comparison and recognition as corroborated through extensive experiments using standard benchmark databases.