Search

In this paper, we investigate properties and limitations of invariance learned by neural networks from the data compared to the genuine invariance achieved through invariant weight-tying. To do so, we adopt a group theoretical perspective and analyze invariance learning in neural networks without weight-tying constraints. We demonstrate that even when a network learns to correctly classify samples on a group orbit, the underlying decision-making in such a model does not attain genuine invariance. Instead, learned invariance is strongly conditioned on the input data, rendering it unreliable if the input distribution shifts. We next demonstrate how to guide invariance learning toward genuine invariance by regularizing the invariance of a model at the training. To this end, we propose several metrics to quantify learned invariance: (i) predictive distribution invariance, (ii) logit invariance, and (iii) saliency invariance similarity. We show that the invariance learned with the invariance error regularization closely reassembles the genuine invariance of weight-tying models and reliably holds even under a severe input distribution shift. Closer analysis of the learned invariance also reveals the spectral decay phenomenon, when a network chooses to achieve the invariance to a specific transformation group by reducing the sensitivity to any input perturbation.

Video summarization aims at choosing parts of a video that narrate a story as close as possible to the original one. Most of the existing video summarization approaches focus on hand-crafted labels. As the number of videos grows exponentially, there emerges an increasing need for methods that can learn meaningful summarizations without labeled annotations. In this paper, we aim to maximally exploit unsupervised video summarization while concentrating the supervision to a few, personalized labels as an add-on. To do so, we formulate the key requirements for the informative video summarization. Then, we propose contrastive learning as the answer to both questions. To further boost Contrastive video Summarization (CSUM), we propose to contrast top-k features instead of a mean video feature as employed by the existing method, which we implement with a differentiable top-k feature selector. Our experiments on several benchmarks demonstrate, that our approach allows for meaningful and diverse summaries when no labeled data is provided.

Symmetries built into a neural network have appeared to be very beneficial for a wide range of tasks as it saves the data to learn them. We depart from the position that when symmetries are not built into a model a priori, it is advantageous for robust networks to learn symmetries directly from the data to fit a task function. In this paper, we present a method to extract symmetries learned by a neural network and to evaluate the degree to which a network is invariant to them. With our method, we are able to explicitly retrieve learned invariances in a form of the generators of corresponding Lie-groups without prior knowledge of symmetries in the data. We use the proposed method to study how symmetrical properties depend on a neural network's parameterization and configuration. We found that the ability of a network to learn symmetries generalizes over a range of architectures. However, the quality of learned symmetries depends on the depth and the number of parameters.

The standard approach to contrastive learning is to maximize the agreement between different views of the data. The views are ordered in pairs, such that they are either positive, encoding different views of the same object, or negative, corresponding to views of different objects. The supervisory signal comes from maximizing the total similarity over positive pairs, while the negative pairs are needed to avoid collapse. In this work, we note that the approach of considering individual pairs cannot account for both intra-set and inter-set similarities when the sets are formed from the views of the data. It thus limits the information content of the supervisory signal available to train representations. We propose to go beyond contrasting individual pairs of objects by focusing on contrasting objects as sets. For this, we use combinatorial quadratic assignment theory designed to evaluate set and graph similarities and derive set-contrastive objective as a regularizer for contrastive learning methods. We conduct experiments and demonstrate that our method improves learned representations for the tasks of metric learning and self-supervised classification.

The goal of this paper is Human-object Interaction (HO-I) detection. HO-I detection aims to find interacting human-objects regions and classify their interaction from an image. Researchers obtain significant improvement in recent years by relying on strong HO-I alignment supervision from [5]. HO-I alignment supervision pairs humans with their interacted objects, and then aligns human-object pair(s) with their interaction categories. Since collecting such annotation is expensive, in this paper, we propose to detect HO-I without alignment supervision. We instead rely on image-level supervision that only enumerates existing interactions within the image without pointing where they happen. Our paper makes three contributions: i) We propose Align-Former, a visual-transformer based CNN that can detect HO-I with only image-level supervision. ii) Align-Former is equipped with HO-I align layer, that can learn to select appropriate targets to allow detector supervision. iii) We evaluate Align-Former on HICO-DET [5] and V-COCO [13], and show that Align-Former outperforms existing image-level supervised HO-I detectors by a large margin (4.71% mAP improvement from 16.14% to 20.85% on HICO-DET [5])., Comment: Accepted at BMVC'21

Robustness against unwanted perturbations is an important aspect of deploying neural network classifiers in the real world. Common natural perturbations include noise, saturation, occlusion, viewpoint changes, and blur deformations. All of them can be modelled by the newly proposed transform-augmented convolutional networks. While many approaches for robustness train the network by providing augmented data to the network, we aim to integrate perturbations in the network architecture to achieve improved and more general robustness. To demonstrate that wiggling the weights consistently improves classification, we choose a standard network and modify it to a transform-augmented network. On perturbed CIFAR-10 images, the modified network delivers a better performance than the original network. For the much smaller STL-10 dataset, in addition to delivering better general robustness, wiggling even improves the classification of unperturbed, clean images substantially. We conclude that wiggled transform-augmented networks acquire good robustness even for perturbations not seen during training., Comment: arXiv admin note: text overlap with arXiv:2103.11372, arXiv:2107.09391

We consider the problem of information compression from high dimensional data. Where many studies consider the problem of compression by non-invertible transformations, we emphasize the importance of invertible compression. We introduce new class of likelihood-based autoencoders with pseudo bijective architecture, which we call Pseudo Invertible Encoders. We provide the theoretical explanation of their principles. We evaluate Gaussian Pseudo Invertible Encoder on MNIST, where our model outperforms WAE and VAE in sharpness of the generated images.

Tracking multiple objects individually differs from tracking groups of related objects. When an object is a part of the group, its trajectory depends on the trajectories of the other group members. Most of the current state-of-the-art trackers follow the approach of tracking each object independently, with the mechanism to handle the overlapping trajectories where necessary. Such an approach does not take inter-object relations into account, which may cause unreliable tracking for the members of the groups, especially in crowded scenarios, where individual cues become unreliable due to occlusions. To overcome these limitations and to extend such trackers to crowded scenes, we propose a plug-in Relation Encoding Module (REM). REM encodes relations between tracked objects by running a message passing over a corresponding spatio-temporal graph, computing relation embeddings for the tracked objects. Our experiments on MOT17 and MOT20 demonstrate that the baseline tracker improves its results after a simple extension with REM. The proposed module allows for tracking severely or even fully occluded objects by utilizing relational cues.

We focus on building robustness in the convolutions of neural visual classifiers, especially against natural perturbations like elastic deformations, occlusions and Gaussian noise. Existing CNNs show outstanding performance on clean images, but fail to tackle naturally occurring perturbations. In this paper, we start from elastic perturbations, which approximate (local) view-point changes of the object. We present elastically-augmented convolutions (EAConv) by parameterizing filters as a combination of fixed elastically-perturbed bases functions and trainable weights for the purpose of integrating unseen viewpoints in the CNN. We show on CIFAR-10 and STL-10 datasets that the general robustness of our method on unseen occlusion, zoom, rotation, image cut and Gaussian perturbations improves, while significantly improving the performance on clean images without any data augmentation.

Scale is often seen as a given, disturbing factor in many vision tasks. When doing so it is one of the factors why we need more data during learning. In recent work scale equivariance was added to convolutional neural networks. It was shown to be effective for a range of tasks. We aim for accurate scale-equivariant convolutional neural networks (SE-CNNs) applicable for problems where high granularity of scale and small kernel sizes are required. Current SE-CNNs rely on weight sharing and kernel rescaling, the latter of which is accurate for integer scales only. To reach accurate scale equivariance, we derive general constraints under which scale-convolution remains equivariant to discrete rescaling. We find the exact solution for all cases where it exists, and compute the approximation for the rest. The discrete scale-convolution pays off, as demonstrated in a new state-of-the-art classification on MNIST-scale and on STL-10 in the supervised learning setting. With the same SE scheme, we also improve the computational effort of a scale-equivariant Siamese tracker on OTB-13.

We focus on the robustness of neural networks for classification. To permit a fair comparison between methods to achieve robustness, we first introduce a standard based on the mensuration of a classifier's degradation. Then, we propose natural perturbed training to robustify the network. Natural perturbations will be encountered in practice: the difference of two images of the same object may be approximated by an elastic deformation (when they have slightly different viewing angles), by occlusions (when they hide differently behind objects), or by saturation, Gaussian noise etc. Training some fraction of the epochs on random versions of such variations will help the classifier to learn better. We conduct extensive experiments on six datasets of varying sizes and granularity. Natural perturbed learning show better and much faster performance than adversarial training on clean, adversarial as well as natural perturbed images. It even improves general robustness on perturbations not seen during the training. For Cifar-10 and STL-10 natural perturbed training even improves the accuracy for clean data and reaches the state of the art performance. Ablation studies verify the effectiveness of natural perturbed training.

This paper studies visual search using structured queries. The structure is in the form of a 2D composition that encodes the position and the category of the objects. The transformation of the position and the category of the objects leads to a continuous-valued relationship between visual compositions, which carries highly beneficial information, although not leveraged by previous techniques. To that end, in this work, our goal is to leverage these continuous relationships by using the notion of symmetry in equivariance. Our model output is trained to change symmetrically with respect to the input transformations, leading to a sensitive feature space. Doing so leads to a highly efficient search technique, as our approach learns from fewer data using a smaller feature space. Experiments on two large-scale benchmarks of MS-COCO and HICO-DET demonstrates that our approach leads to a considerable gain in the performance against competing techniques., Comment: Accepted at WACV'21 (Algorithms Track)

Human-object interaction recognition aims for identifying the relationship between a human subject and an object. Researchers incorporate global scene context into the early layers of deep Convolutional Neural Networks as a solution. They report a significant increase in the performance since generally interactions are correlated with the scene (\ie riding bicycle on the city street). However, this approach leads to the following problems. It increases the network size in the early layers, therefore not efficient. It leads to noisy filter responses when the scene is irrelevant, therefore not accurate. It only leverages scene context whereas human-object interactions offer a multitude of contexts, therefore incomplete. To circumvent these issues, in this work, we propose Self-Selective Context (SSC). SSC operates on the joint appearance of human-objects and context to bring the most discriminative context(s) into play for recognition. We devise novel contextual features that model the locality of human-object interactions and show that SSC can seamlessly integrate with the State-of-the-art interaction recognition models. Our experiments show that SSC leads to an important increase in interaction recognition performance, while using much fewer parameters., Comment: Accepted at ICPR'20

In this paper we aim to explore the general robustness of neural network classifiers by utilizing adversarial as well as natural perturbations. Different from previous works which mainly focus on studying the robustness of neural networks against adversarial perturbations, we also evaluate their robustness on natural perturbations before and after robustification. After standardizing the comparison between adversarial and natural perturbations, we demonstrate that although adversarial training improves the performance of the networks against adversarial perturbations, it leads to drop in the performance for naturally perturbed samples besides clean samples. In contrast, natural perturbations like elastic deformations, occlusions and wave does not only improve the performance against natural perturbations, but also lead to improvement in the performance for the adversarial perturbations. Additionally they do not drop the accuracy on the clean images., Comment: Currently under review

Siamese trackers turn tracking into similarity estimation between a template and the candidate regions in the frame. Mathematically, one of the key ingredients of success of the similarity function is translation equivariance. Non-translation-equivariant architectures induce a positional bias during training, so the location of the target will be hard to recover from the feature space. In real life scenarios, objects undergoe various transformations other than translation, such as rotation or scaling. Unless the model has an internal mechanism to handle them, the similarity may degrade. In this paper, we focus on scaling and we aim to equip the Siamese network with additional built-in scale equivariance to capture the natural variations of the target a priori. We develop the theory for scale-equivariant Siamese trackers, and provide a simple recipe for how to make a wide range of existing trackers scale-equivariant. We present SE-SiamFC, a scale-equivariant variant of SiamFC built according to the recipe. We conduct experiments on OTB and VOT benchmarks and on the synthetically generated T-MNIST and S-MNIST datasets. We demonstrate that a built-in additional scale equivariance is useful for visual object tracking.

Occlusion is one of the most difficult challenges in object tracking to model. This is because unlike other challenges, where data augmentation can be of help, occlusion is hard to simulate as the occluding object can be anything in any shape. In this paper, we propose a simple solution to simulate the effects of occlusion in the latent space. Specifically, we present structured dropout to mimick the change in latent codes under occlusion. We present three forms of dropout (channel dropout, segment dropout and slice dropout) with the various forms of occlusion in mind. To demonstrate its effectiveness, the dropouts are incorporated into two modern Siamese trackers (SiamFC and SiamRPN++). The outputs from multiple dropouts are combined using an encoder network to obtain the final prediction. Experiments on several tracking benchmarks show the benefits of structured dropouts, while due to their simplicity requiring only small changes to the existing tracker models., Comment: Accepted at ICPR2020

Human-object interaction (HOI) detection is a core task in computer vision. The goal is to localize all human-object pairs and recognize their interactions. An interaction defined by a tuple leads to a long-tailed visual recognition challenge since many combinations are rarely represented. The performance of the proposed models is limited especially for the tail categories, but little has been done to understand the reason. To that end, in this paper, we propose to diagnose rarity in HOI detection. We propose a three-step strategy, namely Detection, Identification and Recognition where we carefully analyse the limiting factors by studying state-of-the-art models. Our findings indicate that detection and identification steps are altered by the interaction signals like occlusion and relative location, as a result limiting the recognition accuracy., Comment: Accepted at CVPR'20 Workshop on Learning from Limited Labels

Neural operations as convolutions, self-attention, and vector aggregation are the go-to choices for recognizing short-range actions. However, they have three limitations in modeling long-range activities. This paper presents PIC, Permutation Invariant Convolution, a novel neural layer to model the temporal structure of long-range activities. It has three desirable properties. i. Unlike standard convolution, PIC is invariant to the temporal permutations of features within its receptive field, qualifying it to model the weak temporal structures. ii. Different from vector aggregation, PIC respects local connectivity, enabling it to learn long-range temporal abstractions using cascaded layers. iii. In contrast to self-attention, PIC uses shared weights, making it more capable of detecting the most discriminant visual evidence across long and noisy videos. We study the three properties of PIC and demonstrate its effectiveness in recognizing the long-range activities of Charades, Breakfast, and MultiThumos.

For many of the physical phenomena around us, we have developed sophisticated models explaining their behavior. Nevertheless, measuring physical properties from visual observations is challenging due to the high number of causally underlying physical parameters -- including material properties and external forces. In this paper, we propose to measure latent physical properties for cloth in the wind without ever having seen a real example before. Our solution is an iterative refinement procedure with simulation at its core. The algorithm gradually updates the physical model parameters by running a simulation of the observed phenomenon and comparing the current simulation to a real-world observation. The correspondence is measured using an embedding function that maps physically similar examples to nearby points. We consider a case study of cloth in the wind, with curling flags as our leading example -- a seemingly simple phenomena but physically highly involved. Based on the physics of cloth and its visual manifestation, we propose an instantiation of the embedding function. For this mapping, modeled as a deep network, we introduce a spectral layer that decomposes a video volume into its temporal spectral power and corresponding frequencies. Our experiments demonstrate that the proposed method compares favorably to prior work on the task of measuring cloth material properties and external wind force from a real-world video., Comment: CVPR 2020. arXiv admin note: substantial text overlap with arXiv:1910.07861

In this paper, we aim to explain the decisions of neural networks by utilizing multimodal information. That is counter-intuitive attributes and counter visual examples which appear when perturbed samples are introduced. Different from previous work on interpreting decisions using saliency maps, text, or visual patches we propose to use attributes and counter-attributes, and examples and counter-examples as part of the visual explanations. When humans explain visual decisions they tend to do so by providing attributes and examples. Hence, inspired by the way of human explanations in this paper we provide attribute-based and example-based explanations. Moreover, humans also tend to explain their visual decisions by adding counter-attributes and counter-examples to explain what is not seen. We introduce directed perturbations in the examples to observe which attribute values change when classifying the examples into the counter classes. This delivers intuitive counter-attributes and counter-examples. Our experiments with both coarse and fine-grained datasets show that attributes provide discriminating and human-understandable intuitive and counter-intuitive explanations., Comment: arXiv admin note: substantial text overlap with arXiv:1910.07416, arXiv:1904.08279

For many of the physical phenomena around us, we have developed sophisticated models explaining their behavior. Nevertheless, inferring specifics from visual observations is challenging due to the high number of causally underlying physical parameters -- including material properties and external forces. This paper addresses the problem of inferring such latent physical properties from observations. Our solution is an iterative refinement procedure with simulation at its core. The algorithm gradually updates the physical model parameters by running a simulation of the observed phenomenon and comparing the current simulation to a real-world observation. The physical similarity is computed using an embedding function that maps physically similar examples to nearby points. As a tangible example, we concentrate on flags curling in the wind -- a seemingly simple phenomenon but physically highly involved. Based on its underlying physical model and visual manifestation, we propose an instantiation of the embedding function. For this mapping, modeled as a deep network, we introduce a spectral decomposition layer that decomposes a video volume into its temporal spectral power and corresponding frequencies. In experiments, we demonstrate our method's ability to recover intrinsic and extrinsic physical parameters from both simulated and real-world video.

In this paper, we aim to understand and explain the decisions of deep neural networks by studying the behavior of predicted attributes when adversarial examples are introduced. We study the changes in attributes for clean as well as adversarial images in both standard and adversarially robust networks. We propose a metric to quantify the robustness of an adversarially robust network against adversarial attacks. In a standard network, attributes predicted for adversarial images are consistent with the wrong class, while attributes predicted for the clean images are consistent with the true class. In an adversarially robust network, the attributes predicted for adversarial images classified correctly are consistent with the true class. Finally, we show that the ability to robustify a network varies for different datasets. For the fine grained dataset, it is higher as compared to the coarse-grained dataset. Additionally, the ability to robustify a network increases with the increase in adversarial noise., Comment: arXiv admin note: substantial text overlap with arXiv:1904.08279

The effectiveness of Convolutional Neural Networks (CNNs) has been substantially attributed to their built-in property of translation equivariance. However, CNNs do not have embedded mechanisms to handle other types of transformations. In this work, we pay attention to scale changes, which regularly appear in various tasks due to the changing distances between the objects and the camera. First, we introduce the general theory for building scale-equivariant convolutional networks with steerable filters. We develop scale-convolution and generalize other common blocks to be scale-equivariant. We demonstrate the computational efficiency and numerical stability of the proposed method. We compare the proposed models to the previously developed methods for scale equivariance and local scale invariance. We demonstrate state-of-the-art results on MNIST-scale dataset and on STL-10 dataset in the supervised learning setting.

With the transformative technologies and the rapidly changing global R&D landscape, the multimedia and multimodal community is now faced with many new opportunities and uncertainties. With the open source dissemination platform and pervasive computing resources, new research results are being discovered at an unprecedented pace. In addition, the rapid exchange and influence of ideas across traditional discipline boundaries have made the emphasis on multimedia multimodal research even more important than before. To seize these opportunities and respond to the challenges, we have organized a workshop to specifically address and brainstorm the challenges, opportunities, and research roadmaps for MM research. The two-day workshop, held on March 30 and 31, 2017 in Washington DC, was sponsored by the Information and Intelligent Systems Division of the National Science Foundation of the United States. Twenty-three (23) invited participants were asked to review and identify research areas in the MM field that are most important over the next 10-15 year timeframe. Important topics were selected through discussion and consensus, and then discussed in depth in breakout groups. Breakout groups reported initial discussion results to the whole group, who continued with further extensive deliberation. For each identified topic, a summary was produced after the workshop to describe the main findings, including the state of the art, challenges, and research roadmaps planned for the next 5, 10, and 15 years in the identified area., Comment: Long Report of NSF Workshop on Multimedia Challenges, Opportunities and Research Roadmaps, held in March 2017, Washington DC. Short report available separately

Updating the tracker model with adverse bounding box predictions adds an unavoidable bias term to the learning. This bias term, which we refer to as model decay, offsets the learning and causes tracking drift. While its adverse affect might not be visible in short-term tracking, accumulation of this bias over a long-term can eventually lead to a permanent loss of the target. In this paper, we look at the problem of model bias from a mathematical perspective. Further, we briefly examine the effect of various sources of tracking error on model decay, using a correlation filter (ECO) and a Siamese (SINT) tracker. Based on observations and insights, we propose simple additions that help to reduce model decay in long-term tracking. The proposed tracker is evaluated on four long-term and one short term tracking benchmarks, demonstrating superior accuracy and robustness, even in 30 minute long videos.

Many human activities take minutes to unfold. To represent them, related works opt for statistical pooling, which neglects the temporal structure. Others opt for convolutional methods, as CNN and Non-Local. While successful in learning temporal concepts, they are short of modeling minutes-long temporal dependencies. We propose VideoGraph, a method to achieve the best of two worlds: represent minutes-long human activities and learn their underlying temporal structure. VideoGraph learns a graph-based representation for human activities. The graph, its nodes and edges are learned entirely from video datasets, making VideoGraph applicable to problems without node-level annotation. The result is improvements over related works on benchmarks: Epic-Kitchen and Breakfast. Besides, we demonstrate that VideoGraph is able to learn the temporal structure of human activities in minutes-long videos.

Deep computer vision systems being vulnerable to imperceptible and carefully crafted noise have raised questions regarding the robustness of their decisions. We take a step back and approach this problem from an orthogonal direction. We propose to enable black-box neural networks to justify their reasoning both for clean and for adversarial examples by leveraging attributes, i.e. visually discriminative properties of objects. We rank attributes based on their class relevance, i.e. how the classification decision changes when the input is visually slightly perturbed, as well as image relevance, i.e. how well the attributes can be localized on both clean and perturbed images. We present comprehensive experiments for attribute prediction, adversarial example generation, adversarially robust learning, and their qualitative and quantitative analysis using predicted attributes on three benchmark datasets.

The goal of this paper is to retrieve an image based on instance, attribute and category similarity notions. Different from existing works, which usually address only one of these entities in isolation, we introduce a cooperative embedding to integrate them while preserving their specific level of semantic representation. An algebraic structure defines a superspace filled with instances. Attributes are axis-aligned to form subspaces, while categories influence the arrangement of similar instances. These relationships enable them to cooperate for their mutual benefits for image retrieval. We derive a proxy-based softmax embedding loss to learn simultaneously all similarity measures in both superspace and subspaces. We evaluate our model on datasets from two different domains. Experiments on image retrieval tasks show the benefits of the cooperative embeddings for modeling multiple image similarities, and for discovering style evolution of instances between- and within-categories.

This paper focuses on the temporal aspect for recognizing human activities in videos; an important visual cue that has long been undervalued. We revisit the conventional definition of activity and restrict it to Complex Action: a set of one-actions with a weak temporal pattern that serves a specific purpose. Related works use spatiotemporal 3D convolutions with fixed kernel size, too rigid to capture the varieties in temporal extents of complex actions, and too short for long-range temporal modeling. In contrast, we use multi-scale temporal convolutions, and we reduce the complexity of 3D convolutions. The outcome is Timeception convolution layers, which reasons about minute-long temporal patterns, a factor of 8 longer than best related works. As a result, Timeception achieves impressive accuracy in recognizing the human activities of Charades, Breakfast Actions, and MultiTHUMOS. Further, we demonstrate that Timeception learns long-range temporal dependencies and tolerate temporal extents of complex actions., Comment: IEEE CVPR 2019 (Oral)

Distinction among nearby poses and among symmetries of an object is challenging. In this paper, we propose a unified, group-theoretic approach to tackle both. Different from existing works which directly predict absolute pose, our method measures the pose of an object relative to another pose, i.e., the pose difference. The proposed method generates the complete orbit of an object from a single view of the object with respect to the subgroup of SO(3) of rotations around the z-axis, and compares the orbit of the object with another orbit using a novel orbit metric to estimate the pose difference. The generated orbit in the latent space records all the differences in pose in the original observational space, and as a result, the method is capable of finding subtle differences in pose. We demonstrate the effectiveness of the proposed method on cars, where identifying the subtle pose differences is vital., Comment: to appear in BMVC2018

Visual repetition is ubiquitous in our world. It appears in human activity (sports, cooking), animal behavior (a bee's waggle dance), natural phenomena (leaves in the wind) and in urban environments (flashing lights). Estimating visual repetition from realistic video is challenging as periodic motion is rarely perfectly static and stationary. To better deal with realistic video, we elevate the static and stationary assumptions often made by existing work. Our spatiotemporal filtering approach, established on the theory of periodic motion, effectively handles a wide variety of appearances and requires no learning. Starting from motion in 3D we derive three periodic motion types by decomposition of the motion field into its fundamental components. In addition, three temporal motion continuities emerge from the field's temporal dynamics. For the 2D perception of 3D motion we consider the viewpoint relative to the motion; what follows are 18 cases of recurrent motion perception. To estimate repetition under all circumstances, our theory implies constructing a mixture of differential motion maps: gradient, divergence and curl. We temporally convolve the motion maps with wavelet filters to estimate repetitive dynamics. Our method is able to spatially segment repetitive motion directly from the temporal filter responses densely computed over the motion maps. For experimental verification of our claims, we use our novel dataset for repetition estimation, better-reflecting reality with non-static and non-stationary repetitive motion. On the task of repetition counting, we obtain favorable results compared to a deep learning alternative.

We introduce the OxUvA dataset and benchmark for evaluating single-object tracking algorithms. Benchmarks have enabled great strides in the field of object tracking by defining standardized evaluations on large sets of diverse videos. However, these works have focused exclusively on sequences that are just tens of seconds in length and in which the target is always visible. Consequently, most researchers have designed methods tailored to this "short-term" scenario, which is poorly representative of practitioners' needs. Aiming to address this disparity, we compile a long-term, large-scale tracking dataset of sequences with average length greater than two minutes and with frequent target object disappearance. The OxUvA dataset is much larger than the object tracking datasets of recent years: it comprises 366 sequences spanning 14 hours of video. We assess the performance of several algorithms, considering both the ability to locate the target and to determine whether it is present or absent. Our goal is to offer the community a large and diverse benchmark to enable the design and evaluation of tracking methods ready to be used "in the wild". The project website is http://oxuva.net, Comment: To appear at ECCV 2018

This work incorporates the multi-modality of the data distribution into a Gaussian Process regression model. We approach the problem from a discriminative perspective by learning, jointly over the training data, the target space variance in the neighborhood of a certain sample through metric learning. We start by using data centers rather than all training samples. Subsequently, each center selects an individualized kernel metric. This enables each center to adjust the kernel space in its vicinity in correspondence with the topology of the targets --- a multi-modal approach. We additionally add descriptiveness by allowing each center to learn a precision matrix. We demonstrate empirically the reliability of the model., Comment: Accepted in Pattern Recognition Letters, 2018

This method introduces an efficient manner of learning action categories without the need of feature estimation. The approach starts from low-level values, in a similar style to the successful CNN methods. However, rather than extracting general image features, we learn to predict specific video representations from raw video data. The benefit of such an approach is that at the same computational expense it can predict 2 D video representations as well as 3 D ones, based on motion. The proposed model relies on discriminative Waldboost, which we enhance to a multiclass formulation for the purpose of learning video representations. The suitability of the proposed approach as well as its time efficiency are tested on the UCF11 action recognition dataset., Comment: Published in the proceedings of the International Conference on Image Processing (ICIP), 2016

This paper proposes motion prediction in single still images by learning it from a set of videos. The building assumption is that similar motion is characterized by similar appearance. The proposed method learns local motion patterns given a specific appearance and adds the predicted motion in a number of applications. This work (i) introduces a novel method to predict motion from appearance in a single static image, (ii) to that end, extends of the Structured Random Forest with regression derived from first principles, and (iii) shows the value of adding motion predictions in different tasks such as: weak frame-proposals containing unexpected events, action recognition, motion saliency. Illustrative results indicate that motion prediction is not only feasible, but also provides valuable information for a number of applications., Comment: Published in the proceedings of the European Conference on Computer Vision (ECCV), 2014

We consider the problem of estimating repetition in video, such as performing push-ups, cutting a melon or playing violin. Existing work shows good results under the assumption of static and stationary periodicity. As realistic video is rarely perfectly static and stationary, the often preferred Fourier-based measurements is inapt. Instead, we adopt the wavelet transform to better handle non-static and non-stationary video dynamics. From the flow field and its differentials, we derive three fundamental motion types and three motion continuities of intrinsic periodicity in 3D. On top of this, the 2D perception of 3D periodicity considers two extreme viewpoints. What follows are 18 fundamental cases of recurrent perception in 2D. In practice, to deal with the variety of repetitive appearance, our theory implies measuring time-varying flow and its differentials (gradient, divergence and curl) over segmented foreground motion. For experiments, we introduce the new QUVA Repetition dataset, reflecting reality by including non-static and non-stationary videos. On the task of counting repetitions in video, we obtain favorable results compared to a deep learning alternative.

It is widely believed that the success of deep convolutional networks is based on progressively discarding uninformative variability about the input with respect to the problem at hand. This is supported empirically by the difficulty of recovering images from their hidden representations, in most commonly used network architectures. In this paper we show via a one-to-one mapping that this loss of information is not a necessary condition to learn representations that generalize well on complicated problems, such as ImageNet. Via a cascade of homeomorphic layers, we build the i-RevNet, a network that can be fully inverted up to the final projection onto the classes, i.e. no information is discarded. Building an invertible architecture is difficult, for one, because the local inversion is ill-conditioned, we overcome this by providing an explicit inverse. An analysis of i-RevNets learned representations suggests an alternative explanation for the success of deep networks by a progressive contraction and linear separation with depth. To shed light on the nature of the model learned by the i-RevNet we reconstruct linear interpolations between natural image representations.

Long-term tracking requires extreme stability to the multitude of model updates and robustness to the disappearance and loss of the target as such will inevitably happen. For motivation, we have taken 10 randomly selected OTB-sequences, doubled each by attaching a reversed version and repeated each double sequence 20 times. On most of these repetitive videos, the best current tracker performs worse on each loop. This illustrates the difference between optimization for short-term versus long-term tracking. In a long-term tracker a combined global and local search strategy is beneficial, allowing for recovery from failures and disappearance. Most importantly, the proposed tracker also employs cautious updating, guided by self-quality assessment. The proposed tracker is still among the best on the 20-sec OTB-videos while achieving state-of-the-art on the 100-sec UAV20L benchmark. On 10 new half-an-hour videos with city bicycling, sport games etc, the proposed tracker outperforms others by a large margin where the 2010 TLD tracker comes second., Comment: tech report

Filters in convolutional networks are typically parameterized in a pixel basis, that does not take prior knowledge about the visual world into account. We investigate the generalized notion of frames designed with image properties in mind, as alternatives to this parametrization. We show that frame-based ResNets and Densenets can improve performance on Cifar-10+ consistently, while having additional pleasant properties like steerability. By exploiting these transformation properties explicitly, we arrive at dynamic steerable blocks. They are an extension of residual blocks, that are able to seamlessly transform filters under pre-defined transformations, conditioned on the input at training and inference time. Dynamic steerable blocks learn the degree of invariance from data and locally adapt filters, allowing them to apply a different geometrical variant of the same filter to each location of the feature map. When evaluated on the Berkeley Segmentation contour detection dataset, our approach outperforms all competing approaches that do not utilize pre-training. Our results highlight the benefits of image-based regularization to deep networks.

Event detection in unconstrained videos is conceived as a content-based video retrieval with two modalities: textual and visual. Given a text describing a novel event, the goal is to rank related videos accordingly. This task is zero-exemplar, no video examples are given to the novel event. Related works train a bank of concept detectors on external data sources. These detectors predict confidence scores for test videos, which are ranked and retrieved accordingly. In contrast, we learn a joint space in which the visual and textual representations are embedded. The space casts a novel event as a probability of pre-defined events. Also, it learns to measure the distance between an event and its related videos. Our model is trained end-to-end on publicly available EventNet. When applied to TRECVID Multimedia Event Detection dataset, it outperforms the state-of-the-art by a considerable margin., Comment: IEEE CVPR 2017

Deep neural network algorithms are difficult to analyze because they lack structure allowing to understand the properties of underlying transforms and invariants. Multiscale hierarchical convolutional networks are structured deep convolutional networks where layers are indexed by progressively higher dimensional attributes, which are learned from training data. Each new layer is computed with multidimensional convolutions along spatial and attribute variables. We introduce an efficient implementation of such networks where the dimensionality is progressively reduced by averaging intermediate layers along attribute indices. Hierarchical networks are tested on CIFAR image data bases where they obtain comparable precisions to state of the art networks, with much fewer parameters. We study some properties of the attributes learned from these databases.

This paper explores new evaluation perspectives for image captioning and introduces a noun translation task that achieves comparative image caption generation performance by translating from a set of nouns to captions. This implies that in image captioning, all word categories other than nouns can be evoked by a powerful language model without sacrificing performance on n-gram precision. The paper also investigates lower and upper bounds of how much individual word categories in the captions contribute to the final BLEU score. A large possible improvement exists for nouns, verbs, and prepositions., Comment: This paper was presented at the ECCV2016 2nd Workshop on Storytelling with Images and Videos (VisStory)

In this paper we propose to represent a scene as an abstraction of 'things'. We start from 'things' as generated by modern object proposals, and we investigate their immediately observable properties: position, size, aspect ratio and color, and those only. Where the recent successes and excitement of the field lie in object identification, we represent the scene composition independent of object identities. We make three contributions in this work. First, we study simple observable properties of 'things', and call it things syntax. Second, we propose translating the things syntax in linguistic abstract statements and study their descriptive effect to retrieve scenes. Thirdly, we propose querying of scenes with abstract block illustrations and study their effectiveness to discriminate among different types of scenes. The benefit of abstract statements and block illustrations is that we generate them directly from the images, without any learning beforehand as in the standard attribute learning. Surprisingly, we show that even though we use the simplest of features from 'things' layout and no learning at all, we can still retrieve scenes reasonably well.

This paper aims for generic instance search from one example where the instance can be an arbitrary object like shoes, not just near-planar and one-sided instances like buildings and logos. First, we evaluate state-of-the-art instance search methods on this problem. We observe that what works for buildings loses its generality on shoes. Second, we propose to use automatically learned category-specific attributes to address the large appearance variations present in generic instance search. Searching among instances from the same category as the query, the category-specific attributes outperform existing approaches by a large margin on shoes and cars and perform on par with the state-of-the-art on buildings. Third, we treat person re-identification as a special case of generic instance search. On the popular VIPeR dataset, we reach state-of-the-art performance with the same method. Fourth, we extend our method to search objects without restriction to the specifically known category. We show that the combination of category-level information and the category-specific attributes is superior to the alternative method combining category-level information with low-level features such as Fisher vector., Comment: This technical report is an extended version of our previous conference paper 'Attributes and Categories for Generic Instance Search from One Example' (CVPR 2015)

In this paper we present a tracker, which is radically different from state-of-the-art trackers: we apply no model updating, no occlusion detection, no combination of trackers, no geometric matching, and still deliver state-of-the-art tracking performance, as demonstrated on the popular online tracking benchmark (OTB) and six very challenging YouTube videos. The presented tracker simply matches the initial patch of the target in the first frame with candidates in a new frame and returns the most similar patch by a learned matching function. The strength of the matching function comes from being extensively trained generically, i.e., without any data of the target, using a Siamese deep neural network, which we design for tracking. Once learned, the matching function is used as is, without any adapting, to track previously unseen targets. It turns out that the learned matching function is so powerful that a simple tracker built upon it, coined Siamese INstance search Tracker, SINT, which only uses the original observation of the target from the first frame, suffices to reach state-of-the-art performance. Further, we show the proposed tracker even allows for target re-identification after the target was absent for a complete video shot., Comment: This paper is accepted to the IEEE Conference on Computer Vision and Pattern Recognition, 2016

Learning powerful feature representations with CNNs is hard when training data are limited. Pre-training is one way to overcome this, but it requires large datasets sufficiently similar to the target domain. Another option is to design priors into the model, which can range from tuned hyperparameters to fully engineered representations like Scattering Networks. We combine these ideas into structured receptive field networks, a model which has a fixed filter basis and yet retains the flexibility of CNNs. This flexibility is achieved by expressing receptive fields in CNNs as a weighted sum over a fixed basis which is similar in spirit to Scattering Networks. The key difference is that we learn arbitrary effective filter sets from the basis rather than modeling the filters. This approach explicitly connects classical multiscale image analysis with general CNNs. With structured receptive field networks, we improve considerably over unstructured CNNs for small and medium dataset scenarios as well as over Scattering for large datasets. We validate our findings on ILSVRC2012, Cifar-10, Cifar-100 and MNIST. As a realistic small dataset example, we show state-of-the-art classification results on popular 3D MRI brain-disease datasets where pre-training is difficult due to a lack of large public datasets in a similar domain., Comment: Reason for update: i) Fix Reference for "Deep roto-translation scattering for object classification" by Oyallon and Mallat. ii) Fixed two minor typos. iii) Removed implicit assumption in equation (4) where scale is represented with diffusion time and adapted to rest of paper where scale is represented with standard deviation, to avoid possible confusion