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

Existing 3D face modeling methods usually depend on 3D Morphable Models, which inherently constrain the representation capacity to fixed shape priors. Optimization-based approaches offer high-quality reconstructions but tend to be computationally expensive. In this work, we introduce GLVD, a hybrid method for 3D face reconstruction from few-shot images that extends Learned Vertex Descent (LVD) by integrating per-vertex neural field optimization with global structural guidance from dynamically predicted 3D keypoints. By incorporating relative spatial encoding, GLVD iteratively refines mesh vertices without requiring dense 3D supervision. This enables expressive and adaptable geometry reconstruction while maintaining computational efficiency. GLVD achieves state-of-the-art performance in single-view settings and remains highly competitive in multi-view scenarios, all while substantially reducing inference time.

Recent advancements in learning techniques that employ coordinate-based neural representations have yielded remarkable results in multi-view 3D reconstruction tasks. However, these approaches often require a substantial number of input views (typically several tens) and computationally intensive optimization procedures to achieve their effectiveness. In this paper, we address these limitations specifically for the problem of few-shot full 3D head reconstruction. We accomplish this by incorporating a probabilistic shape and appearance prior into coordinate-based representations, enabling faster convergence and improved generalization when working with only a few input images (even as low as a single image). During testing, we leverage this prior to guiding the fitting process of a signed distance function using a differentiable renderer. By incorporating the statistical prior alongside parallelizable ray tracing and dynamic caching strategies, we achieve an efficient and accurate approach to few-shot full 3D head reconstruction. Moreover, we extend the H3DS dataset, which now comprises 60 high-resolution 3D full-head scans and their corresponding posed images and masks, which we use for evaluation purposes. By leveraging this dataset, we demonstrate the remarkable capabilities of our approach in achieving state-of-the-art results in geometry reconstruction while being an order of magnitude faster than previous approaches.

Natural language plays a critical role in many computer vision applications, such as image captioning, visual question answering, and cross-modal retrieval, to provide fine-grained semantic information. Unfortunately, while human pose is key to human understanding, current 3D human pose datasets lack detailed language descriptions. To address this issue, we have introduced the PoseScript dataset. This dataset pairs more than six thousand 3D human poses from AMASS with rich human-annotated descriptions of the body parts and their spatial relationships. Additionally, to increase the size of the dataset to a scale that is compatible with data-hungry learning algorithms, we have proposed an elaborate captioning process that generates automatic synthetic descriptions in natural language from given 3D keypoints. This process extracts low-level pose information, known as "posecodes", using a set of simple but generic rules on the 3D keypoints. These posecodes are then combined into higher level textual descriptions using syntactic rules. With automatic annotations, the amount of available data significantly scales up (100k), making it possible to effectively pretrain deep models for finetuning on human captions. To showcase the potential of annotated poses, we present three multi-modal learning tasks that utilize the PoseScript dataset. Firstly, we develop a pipeline that maps 3D poses and textual descriptions into a joint embedding space, allowing for cross-modal retrieval of relevant poses from large-scale datasets. Secondly, we establish a baseline for a text-conditioned model generating 3D poses. Thirdly, we present a learned process for generating pose descriptions. These applications demonstrate the versatility and usefulness of annotated poses in various tasks and pave the way for future research in the field.

Recent advances in full-head reconstruction have been obtained by optimizing a neural field through differentiable surface or volume rendering to represent a single scene. While these techniques achieve an unprecedented accuracy, they take several minutes, or even hours, due to the expensive optimization process required. In this work, we introduce InstantAvatar, a method that recovers full-head avatars from few images (down to just one) in a few seconds on commodity hardware. In order to speed up the reconstruction process, we propose a system that combines, for the first time, a voxel-grid neural field representation with a surface renderer. Notably, a naive combination of these two techniques leads to unstable optimizations that do not converge to valid solutions. In order to overcome this limitation, we present a novel statistical model that learns a prior distribution over 3D head signed distance functions using a voxel-grid based architecture. The use of this prior model, in combination with other design choices, results into a system that achieves 3D head reconstructions with comparable accuracy as the state-of-the-art with a 100 $\times$ speed-up.

We present a novel method to generate human motion to populate 3D indoor scenes. It can be controlled with various combinations of conditioning signals such as a path in a scene, target poses, past motions, and scenes represented as 3D point clouds. State-of-the-art methods are either models specialized to one single setting, require vast amounts of high-quality and diverse training data, or are unconditional models that do not integrate scene or other contextual information. As a consequence, they have limited applicability and rely on costly training data. To address these limitations, we propose a new method, dubbed Purposer, based on neural discrete representation learning. Our model is capable of exploiting, in a flexible manner, different types of information already present in open access large-scale datasets such as AMASS. First, we encode unconditional human motion into a discrete latent space. Second, an autoregressive generative model, conditioned with key contextual information, either with prompting or additive tokens, and trained for next-step prediction in this space, synthesizes sequences of latent indices. We further design a novel conditioning block to handle future conditioning information in such a causal model by using a network with two branches to compute separate stacks of features. In this manner, Purposer can generate realistic motion sequences in diverse test scenes. Through exhaustive evaluation, we demonstrate that our multi-contextual solution outperforms existing specialized approaches for specific contextual information, both in terms of quality and diversity. Our model is trained with short sequences, but a byproduct of being able to use various conditioning signals is that at test time different combinations can be used to chain short sequences together and generate long motions within a context scene.

Advances in NERFs have allowed for 3D scene reconstructions and novel view synthesis. Yet, efficiently editing these representations while retaining photo realism is an emerging challenge. Recent methods face three primary limitations: they're slow for interactive use, lack precision at object boundaries, and struggle to ensure multi-view consistency. We introduce IReNe to address these limitations, enabling swift, near real-time color editing in NeRF. Leveraging a pre-trained NeRF model and a single training image with user-applied color edits, IReNe swiftly adjusts network parameters in seconds. This adjustment allows the model to generate new scene views, accurately representing the color changes from the training image while also controlling object boundaries and view-specific effects. Object boundary control is achieved by integrating a trainable segmentation module into the model. The process gains efficiency by retraining only the weights of the last network layer. We observed that neurons in this layer can be classified into those responsible for view-dependent appearance and those contributing to diffuse appearance. We introduce an automated classification approach to identify these neuron types and exclusively fine-tune the weights of the diffuse neurons. This further accelerates training and ensures consistent color edits across different views. A thorough validation on a new dataset, with edited object colors, shows significant quantitative and qualitative advancements over competitors, accelerating speeds by 5x to 500x.

We introduce MultiPhys, a method designed for recovering multi-person motion from monocular videos. Our focus lies in capturing coherent spatial placement between pairs of individuals across varying degrees of engagement. MultiPhys, being physically aware, exhibits robustness to jittering and occlusions, and effectively eliminates penetration issues between the two individuals. We devise a pipeline in which the motion estimated by a kinematic-based method is fed into a physics simulator in an autoregressive manner. We introduce distinct components that enable our model to harness the simulator's properties without compromising the accuracy of the kinematic estimates. This results in final motion estimates that are both kinematically coherent and physically compliant. Extensive evaluations on three challenging datasets characterized by substantial inter-person interaction show that our method significantly reduces errors associated with penetration and foot skating, while performing competitively with the state-of-the-art on motion accuracy and smoothness. Results and code can be found in our project page.

Motion prediction in soccer involves capturing complex dynamics from player and ball interactions. We present FootBots, an encoder-decoder transformer-based architecture addressing motion prediction and conditioned motion prediction through equivariance properties. FootBots captures temporal and social dynamics using set attention blocks and multi-attention block decoder. Our evaluation utilizes two datasets: a real soccer dataset and a tailored synthetic one. Insights from the synthetic dataset highlight the effectiveness of FootBots’ social attention mechanism and the significance of conditioned motion prediction. Empirical results on real soccer data demonstrate that FootBots outperforms baselines in motion prediction and excels in conditioned tasks, such as predicting the players based on the ball position, predicting the offensive (defensive) team based on the ball and the defensive (offensive) team, and predicting the ball position based on all players. Our evaluation connects quantitative and qualitative findings. https://youtu.be/9kaEkfzG3L8

Current methods based on Neural Radiance Fields (NeRF) significantly lack the capacity to quantify uncertainty in their predictions, particularly on the unseen space including the occluded and outside scene content. This limitation hinders their extensive applications in robotics, where the reliability of model predictions has to be considered for tasks such as robotic exploration and planning in unknown environments. To address this, we propose a novel approach to estimate a 3D Uncertainty Field based on the learned incomplete scene geometry, which explicitly identifies these unseen regions. By considering the accumulated transmittance along each camera ray, our Uncertainty Field infers 2D pixel-wise uncertainty, exhibiting high values for rays directly casting towards occluded or outside the scene content. To quantify the uncertainty on the learned surface, we model a stochastic radiance field. Our experiments demonstrate that our approach is the only one that can explicitly reason about high uncertainty both on 3D unseen regions and its involved 2D rendered pixels, compared with recent methods. Furthermore, we illustrate that our designed uncertainty field is ideally suited for real-world robotics tasks, such as next-best-view selection.

In this article, we introduce SMPLicit, a novel generative model to jointly represent body pose, shape and clothing geometry; and LayerNet, a deep network that given a single image of a person simultaneously performs detailed 3D reconstruction of body and clothes. In contrast to existing learning-based approaches that require training specific models for each type of garment, SMPLicit can represent in a unified manner different garment topologies (e.g. from sleeveless tops to hoodies and open jackets), while controlling other properties like garment size or tightness/looseness. LayerNet follows a coarse-to-fine multi-stage strategy by first predicting smooth cloth geometries from SMPLicit, which are then refined by an image-guided displacement network that gracefully fits the body recovering high-frequency details and wrinkles. LayerNet achieves competitive accuracy in the task of 3D reconstruction against current 'garment-agnostic' state of the art for images of people in up-right positions and controlled environments, and consistently surpasses these methods on challenging body poses and uncontrolled settings. Furthermore, the semantically rich outcome of our approach is suitable for performing Virtual Try-on tasks directly on 3D, a task which, so far, has only been addressed in the 2D domain.

While original Neural Radiance Fields (NeRF) have shown impressive results in modeling the appearance of a scene with compact MLP architectures, they are not able to achieve real-time rendering. This has been recently addressed by either baking the outputs of NeRF into a data structure or arranging trainable parameters in an explicit feature grid. These strategies, however, significantly increase the memory footprint of the model which prevents their deployment on bandwidth-constrained applications. In this paper, we extend the grid-based approach to achieve real-time view synthesis at more than 150 FPS using a lightweight model. Our main contribution is a novel architecture in which the density field of NeRF-based representations is split into N regions and the density is modeled using N different decoders which reuse the same feature grid. This results in a smaller grid where each feature is located in more than one spatial position, forcing them to learn a compact representation that is valid for different parts of the scene. We further reduce the size of the final model by disposing of the features symmetrically on each region, which favors feature pruning after training while also allowing smooth gradient transitions between neighboring voxels. An exhaustive evaluation demonstrates that our method achieves real-time performance and quality metrics on a pair with state-of-the-art with an improvement of more than 2× in the FPS/MB ratio.

Automatically producing instructions to modify one’s posture could open the door to endless applications, such as personalized coaching and in-home physical therapy. Tackling the reverse problem (i.e., refining a 3D pose based on some natural language feedback) could help for assisted 3D character animation or robot teaching, for instance. Although a few recent works explore the connections between natural language and 3D human pose, none focus on describing 3D body pose differences. In this paper, we tackle the problem of correcting 3D human poses with natural language. To this end, we introduce the Pose- Fix dataset, which consists of several thousand paired 3D poses and their corresponding text feedback, that describe how the source pose needs to be modified to obtain the target pose. We demonstrate the potential of this dataset on two tasks: (1) text-based pose editing, that aims at generating corrected 3D body poses given a query pose and a text modifier; and (2) correctional text generation, where instructions are generated based on the differences between two body poses. The dataset and the code are available at https://europe.naverlabs.com/research/computer-vision/posefix/.

Recovering the geometry of a human head from a single image, while factorizing the materials and illumination, is a severely ill-posed problem that requires prior information to be solved. Methods based on 3D Morphable Models (3DMM), and their combination with differentiable renderers, have shown promising results. However, the expressiveness of 3DMMs is limited, and they typically yield over-smoothed and identity-agnostic 3D shapes limited to the face region. Highly accurate full head reconstructions have recently been obtained with neural fields that parameterize the geometry using multilayer perceptrons. The versatility of these representations has also proved effective for disentangling geometry, materials and lighting. However, these methods require several tens of input images. In this paper, we introduce SIRA, a method which, from a single image, reconstructs human head avatars with high fidelity geometry and factorized lights and surface materials. Our key ingredients are two data-driven statistical models based on neural fields that resolve the ambiguities of single-view 3D surface reconstruction and appearance factorization. Experiments show that SIRA obtains state of the art results in 3D head reconstruction while at the same time it successfully disentangles the global illumination, and the diffuse and specular albedos. Furthermore, our reconstructions are amenable to physically-based appearance editing and head model relighting.

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

We propose a novel approach to enhance the 3D body pose estimation of a person computed from videos captured from a single wearable camera. The main technical contribution consists of leveraging high-level features linking first- and third-views in a joint embedding space. To learn such embedding space we introduce First2Third-Pose, a new paired synchronized dataset of nearly 2000 videos depicting human activities captured from both first- and third-view perspectives. We explicitly consider spatial- and motion-domain features, combined using a semi-Siamese architecture trained in a self-supervised fashion. Experimental results demonstrate that the joint multi-view embedded space learned with our dataset is useful to extract discriminatory features from arbitrary single-view egocentric videos, with no need to perform any sort of domain adaptation or knowledge of camera parameters. An extensive evaluation demonstrates that we achieve significant improvement in egocentric 3D body pose estimation performance on two unconstrained datasets, over three supervised state-of-the-art approaches. The collected dataset and pre-trained model are available for research purposes.1

In this work, we focus on improving the captions generated by image-caption generation systems. We propose a novel re-ranking approach that leverages visual-semantic measures to identify the ideal caption that maximally captures the visual information in the image. Our re-ranker utilizes the Belief Revision framework (Blok et al., 2003) to calibrate the original likelihood of the top-n captions by explicitly exploiting semantic relatedness between the depicted caption and the visual context. Our experiments demonstrate the utility of our approach, where we observe that our re-ranker can enhance the performance of a typical image-captioning system without necessity of any additional training or fine-tuning.

This paper proposes a do-it-all neural model of human hands, named LISA. The model can capture accurate hand shape and appearance, generalize to arbitrary hand sub-jects, provide dense surface correspondences, be reconstructed from images in the wild, and can be easily an-imated. We train LISA by minimizing the shape and appearance losses on a large set of multi-view RGB image se-quences annotated with coarse 3D poses of the hand skele-ton. For a 3D point in the local hand coordinates, our model predicts the color and the signed distance with respect to each hand bone independently, and then combines the per-bone predictions using the predicted skinning weights. The shape, color, and pose representations are disentangled by design, enabling fine control of the selected hand param-eters. We experimentally demonstrate that LISA can ac-curately reconstruct a dynamic hand from monocular or multi-view sequences, achieving a noticeably higher qual-ity of reconstructed hand shapes compared to baseline approaches. Project page: https://www.iri.upc.edu/people/ecorona/lisa/.

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

As processing power becomes more affordable, computer vision tends to push towards controlling complex processes in the industry. Surface inspection, with changing environmental conditions and the usual lack of homogeneity of the inspected parts, makes it a real challenge to overcome even for a skilled specialist. In addition, the scarcity of positive samples and the extremely small size of the defects, makes it even harder to cluster them in different classes. In this work, we propose a novel training strategy tailored to handle these challenges for the problem of image defect segmentation and classification. First, we propose a Context Aggregation Network with different dilation factors, in order to keep as much information as possible from every feature map, especially for the smallest defects. By splitting the loss in classification and segmentation and positively weighing both terms, we accomplish an optimal learning process counteracting possible imbalances in the dataset. Additionally, we introduce a novel guided-crop image augmentation method, which generates new images by cropping real defects from existing images, pasting them in real non-defective ones and finally tweaking their configuration. This augmentation strategically performed, guided by the evolution of each class loss, allows the model to identify better the least common and complicated to identify defects. We validate our solution with the Magnetic Tile and the Severstal Steel Defect Detection dataset, demonstrating that our approach consistently outperforms models such as ResNet-50, DenseNet-121, HRNet or UPerNet.

Learning controllers that reproduce legged locomotion in nature has been a longtime goal in robotics and computer graphics. While yielding promising results, recent approaches are not yet flexible enough to be applicable to legged systems of different morphologies. This is partly because they often rely on precise motion capture references or elaborate learning environments that ensure the naturality of the emergent locomotion gaits but prevent generalization. This work proposes a generic approach for ensuring realism in locomotion by guiding the learning process with the spring-loaded inverted pendulum model as a reference. Leveraging on the exploration capacities of Reinforcement Learning (RL), we learn a control policy that fills in the information gap between the template model and full-body dynamics required to maintain stable and periodic locomotion. The proposed approach can be applied to robots of different sizes and morphologies and adapted to any RL technique and control architecture. We present experimental results showing that even in a model-free setup and with a simple reactive control architecture, the learned policies can generate realistic and energy-efficient locomotion gaits for a bipedal and a quadrupedal robot. And most importantly, this is achieved without using motion capture, strong constraints in the dynamics or kinematics of the robot, nor prescribing limb coordination. We provide supplemental videos for qualitative analysis of the naturality of the learned gaits4.

We investigated the use of impact sounds generated during exploratory behaviors in a robotic manipulation setup as cues for predicting object surface material and for recognizing individual objects. We collected and make available the YCB-impact sounds dataset which includes over 3,000 impact sounds for the YCB set of everyday objects lying on a table. Impact sounds were generated in three modes: (i) human holding a gripper and hitting, scratching, or dropping the object; (ii) gripper attached to a teleoperated robot hitting the object from the top; (iii) autonomously operated robot hitting the objects from the side with two different speeds. A convolutional neural network is trained from scratch to recognize the object material (steel, aluminium, hard plastic, soft plastic, other plastic, ceramic, wood, paper/cardboard, foam, glass, rubber) from a single impact sound. On the manually collected dataset with more variability in the speed of the action, nearly 60% accuracy for the test set (not presented objects) was achieved. On a robot setup and a stereotypical poking action from top, accuracy of 85% was achieved. This performance drops to 79% if multiple exploratory actions are combined. Individual objects from the set of 75 objects can be recognized with a 79% accuracy. This work demonstrates promising results regarding the possibility of using impact sound for recognition in tasks like single-stream recycling where objects have to be sorted based on their material composition.

This work explores how contextual information and human intention affect the motion prediction of humans during a handover operation with a social robot. By classifying human intention in four different classes, we developed a model able to generate a different motion for each intention class. Furthermore, the model uses a multi-headed attention architecture to add contextual information to the pipeline, such as the position of the robot end effector (REE) or the position of obstacles in the interaction scene. We generate predictions up to two and half seconds in the future given an input sequence of one second containing the previous motion of the human. The results show an improvement of the prediction accuracy, both for the full skeleton prediction and the human hand used for the delivery. The model also allows to generate different sequences with the desired human intention.

In this work we present a novel attention deep learning model that uses context and human intention for 3D human body motion prediction in handover human-robot tasks. This model uses a multi-head attention architecture which incorporates as inputs the human motion, the robot end effector and the position of the obstacles. The outputs of the model are the predicted motion of the human body and the predicted human intention. We use this model to analyze a handover collaborative task with a robot where the robot is able to predict the future motion of the human and use this information in it’s planner. Several experiments are performed where human volunteers fill a standard poll to rate different features, taking into account when the robot uses the prediction versus when the robot doesn’t use the prediction.

3D human pose and shape estimation from monocular images has been an active research area in computer vision. Existing deep learning methods for this task rely on high-resolution input, which however, is not always available in many scenarios such as video surveillance and sports broadcasting. Two common approaches to deal with low-resolution images are applying super-resolution techniques to the input, which may result in unpleasant artifacts, or simply training one model for each resolution, which is impractical in many realistic applications. To address the above issues, this paper proposes a novel algorithm called RSC-Net, which consists of a Resolution-aware network, a Self-supervision loss, and a Contrastive learning scheme. The proposed method is able to learn 3D body pose and shape across different resolutions with one single model. The self-supervision loss enforces scale-consistency of the output, and the contrastive learning scheme enforces scale-consistency of the deep features. We show that both these new losses provide robustness when learning in a weakly-supervised manner. Moreover, we extend the RSC-Net to handle low-resolution videos and apply it to reconstruct textured 3D pedestrians from low-resolution input. Extensive experiments demonstrate that the RSC-Net can achieve consistently better results than the state-of-the-art methods for challenging low-resolution images.

Abstract Given an unordered list of 2D or 3D point trajectories corrupted by noise and partial observations, in this paper we introduce a framework to simultaneously recover the incomplete motion tracks and group the points into spatially and temporally coherent clusters. This advances existing work, which only addresses partial problems and without considering a unified and unsupervised solution. We cast this problem as a matrix completion one, in which point tracks are arranged into a matrix with the missing entries set as zeros. In order to perform the double clustering, the measurement matrix is assumed to be drawn from a dual union of spatiotemporal subspaces. The bases and the dimensionality for these subspaces, the affinity matrices used to encode the temporal and spatial clusters to which each point belongs, and the non-visible tracks, are then jointly estimated via augmented Lagrange multipliers in polynomial time. A thorough evaluation on incomplete motion tracks for multiple-object typologies shows that the accuracy of the matrix we recover compares favorably to that obtained with existing low-rank matrix completion methods, specially under noisy measurements. In addition, besides recovering the incomplete tracks, the point trajectories are directly grouped into different object instances, and a number of semantically meaningful temporal primitive actions are automatically discovered.

We present SIDER (Single-Image neural optimization for facial geometric DEtail Recovery), a novel photometric optimization method that recovers detailed facial geometry from a single image in an unsupervised manner. Inspired by classical techniques of coarse-to-fine optimization and recent advances in implicit neural representations of 3D shape, SIDER combines a geometry prior based on statistical models and Signed Distance Functions (SDFs) to recover facial details from single images. First, it estimates a coarse geometry using a morphable model represented as an SDF. Next, it reconstructs facial geometry details by optimizing a photometric loss with respect to the ground truth image. In contrast to prior work, SIDER does not rely on any dataset priors and does not require additional supervision from multiple views, lighting changes or ground truth 3D shape. Extensive qualitative and quantitative evaluation demonstrates that our method achieves state-of-the-art on facial geometric detail recovery, using only a single in the-wild image.

We introduce PhysXNet, a learning-based approach to predict the dynamics of deformable clothes given 3D skeleton motion sequences of humans wearing these clothes. The proposed model is adaptable to a large variety of garments and changing topologies, without need of being retrained. Such simulations are typically carried out by physics engines that require manual human expertise and are subject to computationally intensive computations. PhysXNet, by contrast, is a fully differentiable deep network that at inference is able to estimate the geometry of dense cloth meshes in a matter of milliseconds, and thus, can be readily deployed as a layer of a larger deep learning architecture. This efficiency is achieved thanks to the specific parameterization of the clothes we consider, based on 3D UV maps encoding spatial garment displacements. The problem is then formulated as a mapping between the human kinematics space (represented also by 3D UV maps of the undressed body mesh) into the clothes displacement UV maps, which we learn using a conditional GAN with a discriminator that enforces feasible deformations. We train simultaneously our model for three garment templates, tops, bottoms and dresses for which we simulate deformations under 50 different human actions. Nevertheless, the UV map representation we consider allows encapsulating many different cloth topologies, and at test we can simulate garments even if we did not specifically train for them. A thorough evaluation demonstrates that PhysXNet delivers cloth deformations very close to those computed with the physical engine, opening the door to be effectively integrated within deep learning pipelines.

Neural Radiance Fields (NeRF) has become a popular framework for learning implicit 3D representations and addressing different tasks such as novel-view synthesis or depth-map estimation. However, in downstream applications where decisions need to be made based on automatic predictions, it is critical to leverage the confidence associated with the model estimations. Whereas uncertainty quantification is a long-standing problem in Machine Learning, it has been largely overlooked in the recent NeRF literature. In this context, we propose Stochastic Neural Radiance Fields (S-NeRF), a generalization of standard NeRF that learns a probability distribution over all the possible radiance fields modeling the scene. This distribution allows to quantify the uncertainty associated with the scene information provided by the model. S-NeRF optimization is posed as a Bayesian learning problem that is efficiently addressed using the Variational Inference framework. Exhaustive experiments over benchmark datasets demonstrate that S-NeRF is able to provide more reliable predictions and confidence values than generic approaches previously proposed for uncertainty estimation in other domains.

We address the problem of multi-person 3D body pose and shape estimation from a single image. While this problem can be addressed by applying single-person approaches multiple times for the same scene, recent works have shown the advantages of building upon deep architectures that simultaneously reason about all people in the scene in a holistic manner by enforcing, e.g., depth order constraints or minimizing interpenetration among reconstructed bodies. However, existing approaches are still unable to capture the size variability of people caused by the inherent body scale and depth ambiguity. In this work, we tackle this challenge by devising a novel optimization scheme that learns the appropriate body scale and relative camera pose, by enforcing the feet of all people to remain on the ground floor. A thorough evaluation on MuPoTS- 3D and 3DPW datasets demonstrates that our approach is able to robustly estimate the body translation and shape of multiple people while retrieving their spatial arrangement, consistently improving current state-of-the-art, especially in scenes with people of very different heights. Code can be found at: https://github.com/nicolasugrinovic/size_depth_disambiguation

In this paper we introduce SMPLicit, a novel generative model to jointly represent body pose, shape and clothing geometry. In contrast to existing learning-based approaches that require training specific models for each type of garment, SMPLicit can represent in a unified manner different garment topologies (e.g. from sleeveless tops to hoodies and to open jackets), while controlling other properties like the garment size or tightness/looseness. We show our model to be applicable to a large variety of garments including T-shirts, hoodies, jackets, shorts, pants, skirts, shoes and even hair. The representation flexibility of SMPLicit builds upon an implicit model conditioned with the SMPL human body parameters and a learnable latent space which is semantically interpretable and aligned with the clothing attributes. The proposed model is fully differentiable, allowing for its use into larger end-to-end trainable systems. In the experimental section, we demonstrate SMPLicit can be readily used for fitting 3D scans and for 3D reconstruction in images of dressed people. In both cases we are able to go beyond state of the art, by retrieving complex garment geometries, handling situations with multiple clothing layers and providing a tool for easy outfit editing. To stimulate further research in this direction, we will make our code and model publicly available at http://www.iri.upc.edu/people/ecorona/smplicit/.

Neural rendering techniques combining machine learning with geometric reasoning have arisen as one of the most promising approaches for synthesizing novel views of a scene from a sparse set of images. Among these, stands out the Neural radiance fields (NeRF) [31], which trains a deep network to map 5D input coordinates (representing spatial location and viewing direction) into a volume density and view-dependent emitted radiance. However, despite achieving an unprecedented level of photorealism on the generated images, NeRF is only applicable to static scenes, where the same spatial location can be queried from different images. In this paper we introduce D-NeRF, a method that extends neural radiance fields to a dynamic domain, allowing to reconstruct and render novel images of objects under rigid and non-rigid motions from a single camera moving around the scene. For this purpose we consider time as an additional input to the system, and split the learning process in two main stages: one that encodes the scene into a canonical space and another that maps this canonical representation into the deformed scene at a particular time. Both mappings are simultaneously learned using fully-connected networks. Once the networks are trained, D-NeRF can render novel images, controlling both the camera view and the time variable, and thus, the object movement. We demonstrate the effectiveness of our approach on scenes with objects under rigid, articulated and non-rigid motions. Code, model weights and the dynamic scenes dataset will be available at [1].

Very recently, the Neural Cellular Automata (NCA) has been proposed to simulate the morphogenesis process with deep networks. NCA learns to grow an image starting from a fixed single pixel. In this work, we show that the neural network (NN) architecture of the NCA can be encapsulated in a larger NN. This allows us to propose a new model that encodes a manifold of NCA, each of them capable of generating a distinct image. Therefore, we are effectively learning an embedding space of CA, which shows generalization capabilities. We accomplish this by introducing dynamic convolutions inside an Auto-Encoder architecture, for the first time used to join two different sources of information, the encoding and cell’s environment information. In biological terms, our approach would play the role of the transcription factors, modulating the mapping of genes into specific proteins that drive cellular differentiation, which occurs right before the morphogenesis. We thoroughly evaluate our approach in a dataset of synthetic emojis and also in real images of CIFAR-10. Our model introduces a general-purpose network, which can be used in a broad range of problems beyond image generation.

Perspective-n-Point-and-Line (PnPL) algorithms aim at fast, accurate, and robust camera localization with respect to a 3D model from 2D-3D feature correspondences, being a major part of modern robotic and AR/VR systems. Current point-based pose estimation methods use only 2D feature detection uncertainties, and the line-based methods do not take uncertainties into account. In our setup, both 3D co-ordinates and 2D projections of the features are considered uncertain. We propose PnP(L) solvers based on EPnP [20] and DLS [14] for the uncertainty-aware pose estimation. We also modify motion-only bundle adjustment to take 3D uncertainties into account. We perform exhaustive synthetic and real experiments on two different visual odometry datasets. The new PnP(L) methods outperform the state-of-the-art on real data in isolation, showing an increase in mean translation accuracy by 18% on a representative subset of KITTI, while the new uncertain refinement improves pose accuracy for most of the solvers, e.g. decreasing mean translation error for the EPnP by 16% compared to the standard refinement on the same dataset. The code is available at https://alexandervakhitov.github.io/uncertain-pnp/.

Recent learning approaches that implicitly represent surface geometry using coordinate-based neural representations have shown impressive results in the problem of multi-view 3D reconstruction. The effectiveness of these techniques is, however, subject to the availability of a large number (several tens) of input views of the scene, and computationally demanding optimizations. In this paper, we tackle these limitations for the specific problem of few-shot full 3D head reconstruction, by endowing coordinate-based representations with a probabilistic shape prior that enables faster convergence and better generalization when using few input images (down to three). First, we learn a shape model of 3D heads from thousands of incomplete raw scans using implicit representations. At test time, we jointly overfit two coordinate-based neural networks to the scene, one modelling the geometry and another estimating the surface radiance, using implicit differentiable rendering. We devise a two-stage optimization strategy in which the learned prior is used to initialize and constrain the geometry during an initial optimization phase. Then, the prior is unfrozen and fine-tuned to the scene. By doing this, we achieve high-fidelity head reconstructions, including hair and shoulders, and with a high level of detail that consistently outperforms both state-of-the-art 3D Morphable Models methods in the few-shot scenario, and nonparametric methods when large sets of views are available.

Attribution map visualization has arisen as one of the most effective techniques to understand the underlying inference process of Convolutional Neural Networks. In this task, the goal is to compute an score for each image pixel related to its contribution to the network output. In this paper, we introduce Disentangled Masked Backpropagation (DMBP), a novel gradient-based method that leverages on the piecewise linear nature of ReLU networks to decompose the model function into different linear mappings. This decomposition aims to disentangle the attribution maps into positive, negative and nuisance factors by learning a set of variables masking the contribution of each filter during back-propagation. A thorough evaluation over standard architectures (ResNet50 and VGG16) and benchmark datasets (PASCAL VOC and ImageNet) demonstrates that DMBP generates more visually interpretable attribution maps than previous approaches. Additionally, we quantitatively show that the maps produced by our method are more consistent with the true contribution of each pixel to the final network output.

While there exists many methods for manipulating rigid objects with parallel-jaw grippers, grasping with multi-finger robotic hands remains a quite unexplored research topic. Reasoning and planning collision-free trajectories on the additional degrees of freedom of several fingers represents an important challenge that, so far, involves computationally costly and slow processes. In this work, we present Multi-FinGAN, a fast generative multi-finger grasp sampling method that synthesizes high quality grasps directly from RGB-D images in about a second. We achieve this by training in an end-to-end fashion a coarse-to-fine model composed of a classification network that distinguishes grasp types according to a specific taxonomy and a refinement network that produces refined grasp poses and joint angles. We experimentally validate and benchmark our method against a standard grasp-sampling method on 790 grasps in simulation and 20 grasps on a real Franka Emika Panda. All experimental results using our method show consistent improvements both in terms of grasp quality metrics and grasp success rate. Remarkably, our approach is up to 20-30 times faster than the baseline, a significant improvement that opens the door to feedback-based grasp re-planning and task informative grasping. Code is available at https://irobotics.aalto.fi/multi-fingan/.

This work proposes a human motion prediction model for handover operations. We use in this work, the different phases of the handover operation to improve the human motion predictions. Our attention deep learning based model takes into account the position of the robot’s End Effector (REE) and the phase in the handover operation to predict future human poses. Our model outputs a distribution of possible positions rather than one deterministic position, a key feature in order to allow robots to collaborate with humans. We provide results of the human upper body and the human right hand, also referred as Human End Effector (HEE).The attention deep learning based model has been trained and evaluated with a dataset created using human volunteers and an anthropomorphic robot, simulating handover operations where the robot is the giver and the human the receiver. For each operation, the human skeleton is obtained with an Intel RealSense D435i camera attached inside the robot’s head. The results shown a great improvement of the human’s right hand prediction and 3D body compared with other methods.

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

We present a novel vehicle detection and tracking system that works solely on 3D LiDAR information. Our approach segments vehicles using a dual-view representation of the 3D LiDAR point cloud on two independently trained convolutional neural networks, one for each view. A bounding box growing algorithm is applied to the fused output of the networks to properly enclose the segmented vehicles. Bounding boxes are grown using a probabilistic method that takes into account also occluded areas. The final vehicle bounding boxes act as observations for a multi-hypothesis tracking system which allows to estimate the position and velocity of the observed vehicles. We thoroughly evaluate our system on the KITTI benchmarks both for detection and tracking separately and show that our dual-branch classifier consistently outperforms previous single-branch approaches, improving or directly competing to other state of the art LiDAR-based methods.

The problem of predicting human motion given a sequence of past observations is at the core of many applications in robotics and computer vision. Current state-of-the-art formulate this problem as a sequence-to-sequence task, in which a historical of 3D skeletons feeds a Recurrent Neural Network (RNN) that predicts future movements, typically in the order of 1 to 2 seconds. However, one aspect that has been obviated so far, is the fact that human motion is inherently driven by interactions with objects and/or other humans in the environment. In this paper, we explore this scenario using a novel context-aware motion prediction architecture. We use a semantic-graph model where the nodes parameterize the human and objects in the scene and the edges their mutual interactions. These interactions are iteratively learned through a graph attention layer, fed with the past observations, which now include both object and human body motions. Once this semantic graph is learned, we inject it to a standard RNN to predict future movements of the human/s and object/s. We consider two variants of our architecture, either freezing the contextual interactions in the future of updating them. A thorough evaluation in the “Whole-Body Human Motion Database” [29] shows that in both cases, our context-aware networks clearly outperform baselines in which the context information is not considered.

The rise of deep learning has brought remarkable progress in estimating hand geometry from images where the hands are part of the scene. This paper focuses on a new problem not explored so far, consisting in predicting how a human would grasp one or several objects, given a single RGB image of these objects. This is a problem with enormous potential in e.g. augmented reality, robotics or prosthetic design. In order to predict feasible grasps, we need to understand the semantic content of the image, its geometric structure and all potential interactions with a hand physical model. To this end, we introduce a generative model that jointly reasons in all these levels and 1) regresses the 3D shape and pose of the objects in the scene; 2) estimates the grasp types; and 3) refines the 51-DoF of a 3D hand model that minimize a graspability loss. To train this model we build the YCB-Affordance dataset, that contains more than 133k images of 21 objects in the YCB-Video dataset. We have annotated these images with more than 28M plausible 3D human grasps according to a 33-class taxonomy. A thorough evaluation in synthetic and real images shows that our model can robustly predict realistic grasps, even in cluttered scenes with multiple objects in close contact.

Flow-based generative models have highly desirable properties like exact log-likelihood evaluation and exact latent-variable inference, however they are still in their infancy and have not received as much attention as alternative generative models. In this paper, we introduce C-Flow, a novel conditioning scheme that brings normalizing flows to an entirely new scenario with great possibilities for multimodal data modeling. C-Flow is based on a parallel sequence of invertible mappings in which a source flow guides the target flow at every step, enabling fine-grained control over the generation process. We also devise a new strategy to model unordered 3D point clouds that, in combination with the conditioning scheme, makes it possible to address 3D reconstruction from a single image and its inverse problem of rendering an image given a point cloud. We demonstrate our conditioning method to be very adaptable, being also applicable to image manipulation, style transfer and multi-modal image-to-image mapping in a diversity of domains, including RGB images, segmentation maps and edge masks.

Designing optimal and light weight networks to fit in resource-limited platforms like mobiles, DSPs or GPUs is a challenging problem with a wide range of interesting applications, e.g. in embedded systems for autonomous driving. While most approaches are based on manual hyperparameter tuning, there exist a new line of research, the so-called NAS (Neural Architecture Search) methods, that aim to optimize several metrics during the design process, including memory requirements of the network, number of FLOPs, number of MACs (Multiply-ACcumulate operations) or inference latency. However, while NAS methods have shown very promising results, they are still significantly time and cost consuming. In this work we introduce E-DNAS, a differentiable architecture search method, which improves the efficiency of NAS methods in designing light-weight networks for the task of image classification. Concretely, E-DNAS computes, in a differentiable manner, the optimal size of a number of meta-kernels that capture patterns of the input data at different resolutions. We also leverage on the additive property of convolution operations to merge several kernels with different compatible sizes into a single one, reducing thus the number of operations and the time required to estimate the optimal configuration. We evaluate our approach on several datasets to perform classification. We report results in terms of the SoC (System on Chips) metric, typically used in the Texas Instruments TDA2x families for autonomous driving applications. The results show that our approach allows designing low latency architectures significantly faster than state-of-the-art.

Ahmed Sabir, Francesc Moreno, Lluís Padró. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Recent advances in 3D human shape estimation build upon parametric representations that model very well the shape of the naked body, but are not appropriate to represent the clothing geometry. In this paper, we present an approach to model dressed humans and predict their geometry from single images. We contribute in three fundamental aspects of the problem, namely, a new dataset, a novel shape parameterization algorithm and an end-to-end deep generative network for predicting shape. First, we present 3D-People, a large-scale synthetic dataset with 2 Million photo-realistic images of 80 subjects performing 70 activities and wearing diverse outfits. Besides providing textured 3D meshes for clothes and body we annotated the dataset with segmentation masks, skeletons, depth, normal maps and optical flow. All this together makes 3D-People suitable for a plethora of tasks. We then represent the 3D shapes using 2D geometry images. To build these images we propose a novel spherical area-preserving parameterization algorithm based on the optimal mass transportation method. We show this approach to improve existing spherical maps which tend to shrink the elongated parts of the full body models such as the arms and legs, making the geometry images incomplete. Finally, we design a multi-resolution deep generative network that, given an input image of a dressed human, predicts his/her geometry image (and thus the clothed body shape) in an end-to-end manner. We obtain very promising results in jointly capturing body pose and clothing shape, both for synthetic validation and on the wild images.

We propose a Generative Adversarial Network (GAN) to forecast 3D human motion given a sequence of past 3D skeleton poses. While recent GANs have shown promising results, they can only forecast plausible motion over relatively short periods of time (few hundred milliseconds) and typically ignore the absolute position of the skeleton w.r.t. the camera. Our scheme provides long term predictions (two seconds or more) for both the body pose and its absolute position. Our approach builds upon three main contributions. First, we represent the data using a spatio-temporal tensor of 3D skeleton coordinates which allows formulating the prediction problem as an inpainting one, for which GANs work particularly well. Secondly, we design an architecture to learn the joint distribution of body poses and global motion, capable to hypothesize large chunks of the input 3D tensor with missing data. And finally, we argue that the L2 metric, considered so far by most approaches, fails to capture the actual distribution of long-term human motion. We propose two alternative metrics, based on the distribution of frequencies, that are able to capture more realistic motion patterns. Extensive experiments demonstrate our approach to significantly improve the state of the art, while also handling situations in which past observations are corrupted by occlusions, noise and missing frames.

In this paper we present an approach to reconstruct the 3D shape of multiple deforming objects from a collection of sparse, noisy and possibly incomplete 2D point tracks acquired by a single monocular camera. Additionally, the proposed solution estimates the camera motion and reasons about the spatial segmentation (i.e., identifies each of the deforming objects in every frame) and temporal clustering (i.e., splits the sequence into motion primitive actions). This advances competing work, which mainly tackled the problem for one single object and non-occluded tracks. In order to handle several objects at a time from partial observations, we model point trajectories as a union of spatial and temporal subspaces, and optimize the parameters of both modalities, the non-observed point tracks, the camera motion, and the time-varying 3D shape via augmented Lagrange multipliers. The algorithm is fully unsupervised and does not require any training data at all. We thoroughly validate the method on challenging scenarios with several human subjects performing different activities which involve complex motions and close interaction. We show our approach achieves state-of-the-art 3D reconstruction results, while it also provides spatial and temporal segmentation.

In this paper, we propose a novel interpretable shape model to encode object nonrigidity. We first use the initial frames of a monocular video to recover a rest shape, used later to compute a dissimilarity measure based on a distance matrix measurement. Spectral analysis is then applied to this matrix to obtain a reduced shape basis, that in contrast to existing approaches, can be physically interpreted. In turn, these precomputed shape bases are used to linearly span the deformation of a wide variety of objects. We introduce the low-rank basis into a sequential approach to recover both camera motion and nonrigid shape from the monocular video, by simply optimizing the weights of the linear combination using bundle adjustment. Since the number of parameters to optimize per frame is relatively small, specially when physical priors are considered, our approach is fast and can potentially run in real time. Validation is done in a wide variety of real-world objects, undergoing both inextensible and extensible deformations. Our approach achieves remarkable robustness to artifacts such as noisy and missing measurements and shows an improved performance to competing methods.

This paper introduces an approach to simultaneously estimate 3D shape, camera pose, and object and type of deformation clustering, from partial 2D annotations in a multi-instance collection of images. Furthermore, we can indistinctly process rigid and non-rigid categories. This advances existing work, which only addresses the problem for one single object or, if multiple objects are considered, they are assumed to be clustered a priori. To handle this broader version of the problem, we model object deformation using a formulation based on multiple unions of subspaces, able to span from small rigid motion to complex deformations. The parameters of this model are learned via Augmented Lagrange Multipliers, in a completely unsupervised manner that does not require any training data at all. Extensive validation is provided in a wide variety of synthetic and real scenarios, including rigid and non-rigid categories with small and large deformations. In all cases our approach outperforms state-of-the-art in terms of 3D reconstruction accuracy, while also providing clustering results that allow segmenting the images into object instances and their associated type of deformation (or action the object is performing).

We propose a method for predicting the 3D shape of a deformable surface from a single view. By contrast with previous approaches, we do not need a pre-registered template of the surface, and our method is robust to the lack of texture and partial occlusions. At the core of our approach is a geometry-aware deep architecture that tackles the problem as usually done in analytic solutions: first perform 2D detection of the mesh and then estimate a 3D shape that is geometrically consistent with the image. We train this architecture in an end-to-end manner using a large dataset of synthetic renderings of shapes under different levels of deformation, material properties, textures and lighting conditions. We evaluate our approach on a test split of this dataset and available real benchmarks, consistently improving state-of-the-art solutions with a significantly lower computational time.

We present a novel approach for synthesizing photorealistic images of people in arbitrary poses using generative adversarial learning. Given an input image of a person and a desired pose represented by a 2D skeleton, our model renders the image of the same person under the new pose, synthesizing novel views of the parts visible in the input image and hallucinating those that are not seen. This problem has recently been addressed in a supervised manner [16, 35], i.e., during training the ground truth images under the new poses are given to the network. We go beyond these approaches by proposing a fully unsupervised strategy. We tackle this challenging scenario by splitting the problem into two principal subtasks. First, we consider a pose conditioned bidirectional generator that maps back the initially rendered image to the original pose, hence being directly comparable to the input image without the need to resort to any training image. Second, we devise a novel loss function that incorporates content and style terms, and aims at producing images of high perceptual quality. Extensive experiments conducted on the DeepFashion dataset demonstrate that the images rendered by our model are very close in appearance to those obtained by fully supervised approaches.

This paper presents an approach to jointly retrieve camera pose, time-varying 3D shape, and automatic clustering based on motion primitives, from incomplete 2D trajectories in a monocular video. We introduce the concept of order-varying temporal regularization in order to exploit video data, that can be indistinctly applied to the 3D shape evolution as well as to the similarities between images. This results in a union of regularized subspaces which effectively encodes the 3D shape deformation. All parameters are learned via augmented Lagrange multipliers, in a unified and unsupervised manner that does not assume any training data at all. Experimental validation is reported on human motion from sparse to dense shapes, providing more robust and accurate solutions than state-of-the-art approaches in terms of 3D reconstruction, while also obtaining motion grouping results.

Automatically changing the expression and physical features of a face from an input image is a topic that has been traditionally tackled in a 2D domain. In this paper, we bring this problem to 3D and propose a framework that given an input RGB video of a human face under a neutral expression, initially computes his/her 3D shape and then performs a transfer to a new and potentially non-observed expression. For this purpose, we parameterize the rest shape -obtained from standard factorization approaches over the input video- using a triangular mesh which is further clustered into larger macro-segments. The expression transfer problem is then posed as a direct mapping between this shape and a source shape, such as the blend shapes of an off-the-shelf 3D dataset of human facial expressions. The mapping is resolved to be geometrically consistent between 3D models by requiring points in specific regions to map on semantic equivalent regions. We validate the approach on several synthetic and real examples of input faces that largely differ from the source shapes, yielding very realistic expression transfers even in cases with topology changes, such as a synthetic video sequence of a single-eyed cyclops.

In this paper we propose a novel approach to estimate dense optical flow from sparse lidar data acquired on an autonomous vehicle. This is intended to be used as a drop-in replacement of any image-based optical flow system when images are not reliable due to e.g. adverse weather conditions or at night. In order to infer high resolution 2D flows from discrete range data we devise a three-block architecture of multiscale filters that combines multiple intermediate objectives, both in the lidar and image domain. To train this network we introduce a dataset with approximately 20K lidar samples of the Kitti dataset which we have augmented with a pseudo ground-truth image-based optical flow computed using FlowNet2. We demonstrate the effectiveness of our approach on Kitti, and show that despite using the low-resolution and sparse measurements of the lidar, we can regress dense optical flow maps which are at par with those estimated with image-based methods.

Perception technologies in Autonomous Driving are experiencing their golden age due to the advances in Deep Learning. Yet, most of these systems rely on the semantically rich information of RGB images. Deep Learning solutions applied to the data of other sensors typically mounted on autonomous cars (e.g. lidars or radars) are not explored much. In this paper we propose a novel solution to understand the dynamics of moving vehicles of the scene from only lidar information. The main challenge of this problem stems from the fact that we need to disambiguate the proprio-motion of the “observer” vehicle from that of the external “observed” vehicles. For this purpose, we devise a CNN architecture which at testing time is fed with pairs of consecutive lidar scans. However, in order to properly learn the parameters of this network, during training we introduce a series of so-called pretext tasks which also leverage on image data. These tasks include semantic information about vehicleness and a novel lidar-flow feature which combines standard image-based optical flow with lidar scans. We obtain very promising results and show that including distilled image information only during training, allows improving the inference results of the network at test time, even when image data is no longer used.

This paper addresses the problem of simultaneously recovering 3D shape, pose and the elastic model of a deformable object from only 2D point tracks in a monocular video. This is a severely under-constrained problem that has been typically addressed by enforcing the shape or the point trajectories to lie on low-rank dimensional spaces. We show that formulating the problem in terms of a low-rank force space that induces the deformation and introducing the elastic model as an additional unknown, allows for a better physical interpretation of the resulting priors and a more accurate representation of the actual object's behavior. In order to simultaneously estimate force, pose, and the elastic model of the object we use an expectation maximization strategy, where each of these parameters are successively learned by partial M-steps. Once the elastic model is learned, it can be transfered to similar objects to code its 3D deformation. Moreover, our approach can robustly deal with missing data, and encode both rigid and non-rigid points under the same formalism. We thoroughly validate the approach on Mocap and real sequences, showing more accurate 3D reconstructions than state-of-the-art, and additionally providing an estimate of the full elastic model with no a priori information.

Building upon recent Deep Neural Network architectures, current approaches lying in the intersection of Computer Vision and Natural Language Processing have achieved unprecedented breakthroughs in tasks like automatic captioning or image retrieval. Most of these learning methods, though, rely on large training sets of images associated with human annotations that specifically describe the visual content. In this paper we propose to go a step further and explore the more complex cases where textual descriptions are loosely related to the images. We focus on the particular domain of news articles in which the textual content often expresses connotative and ambiguous relations that are only suggested but not directly inferred from images. We introduce an adaptive CNN architecture that shares most of the structure for multiple tasks including source detection, article illustration and geolocation of articles. Deep Canonical Correlation Analysis is deployed for article illustration, and a new loss function based on Great Circle Distance is proposed for geolocation. Furthermore, we present BreakingNews, a novel dataset with approximately 100K news articles including images, text and captions, and enriched with heterogeneous meta-data (such as GPS coordinates and user comments). We show this dataset to be appropriate to explore all aforementioned problems, for which we provide a baseline performance using various Deep Learning architectures, and different representations of the textual and visual features. We report very promising results and bring to light several limitations of current state-of-the-art in this kind of domain, which we hope will help spur progress in the field.

In this paper we introduce the Boosted Random Ferns (BRFs) to rapidly build discriminative classifiers for learning and detecting object categories. At the core of our approach we use standard random ferns, but we introduce four main innovations that let us bring ferns from an instance to a category level, and still retain efficiency. First, we define binary features on the histogram of oriented gradients-domain (as opposed to intensity-), allowing for a better representation of intra-class variability. Second, both the positions where ferns are evaluated within the sliding window, and the location of the binary features for each fern are not chosen completely at random, but instead we use a boosting strategy to pick the most discriminative combination of them. This is further enhanced by our third contribution, that is to adapt the boosting strategy to enable sharing of binary features among different ferns, yielding high recognition rates at a low computational cost. And finally, we show that training can be performed online, for sequentially arriving images. Overall, the resulting classifier can be very efficiently trained, densely evaluated for all image locations in about 0.1 seconds, and provides detection rates similar to competing approaches that require expensive and significantly slower processing times. We demonstrate the effectiveness of our approach by thorough experimentation in publicly available datasets in which we compare against state-of-the-art, and for tasks of both 2D detection and 3D multi-view estimation.

We present an approach to reconstruct the 3D shape of multiple deforming objects from incomplete 2D trajectories acquired by a single camera. Additionally, we simultaneously provide spatial segmentation (i.e., we identify each of the objects in every frame) and temporal clustering (i.e., we split the sequence into primitive actions). This advances existing work, which only tackled the problem for one single object and non-occluded tracks. In order to handle several objects at a time from partial observations, we model point trajectories as a union of spatial and temporal subspaces, and optimize the parameters of both modalities, the non-observed point tracks and the 3D shape via augmented Lagrange multipliers. The algorithm is fully unsupervised and results in a formulation which does not need initialization. We thoroughly validate the method on challenging scenarios with several human subjects performing different activities which involve complex motions and close interaction. We show our approach achieves state-of-the-art 3D reconstruction results, while it also provides spatial and temporal segmentation.

This paper addresses the problem of 3D human pose estimation from a single image. We follow a standard two-step pipeline by first detecting the 2D position of the N body joints, and then using these observations to infer 3D pose. For the first step, we use a recent CNN-based detector. For the second step, most existing approaches perform 2N-to-3N regression of the Cartesian joint coordinates. We show that more precise pose estimates can be obtained by representing both the 2D and 3D human poses using NxN distance matrices, and formulating the problem as a 2D-to-3D distance matrix regression. For learning such a regressor we leverage on simple Neural Network architectures, which by construction, enforce positivity and symmetry of the predicted matrices. The approach has also the advantage to naturally handle missing observations and allowing to hypothesize the position of non-observed joints. Quantitative results on Humaneva and Human3.6M datasets demonstrate consistent performance gains over state-of-the-art. Qualitative evaluation on the images in-the-wild of the LSP dataset, using the regressor learned on Human3.6M, reveals very promising generalization results.

Finding a product in the fashion world can be a daunting task. Everyday, e-commerce sites are updating with thousands of images and their associated metadata (textual information), deepening the problem, akin to finding a needle in a haystack. In this paper, we leverage both the images and textual metadata and propose a joint multi-modal embedding that maps both the text and images into a common latent space. Distances in the latent space correspond to similarity between products, allowing us to effectively perform retrieval in this latent space, which is both efficient and accurate. We train this embedding using large-scale real world e-commerce data by both minimizing the similarity between related products and using auxiliary classification networks to that encourage the embedding to have semantic meaning. We compare against existing approaches and show significant improvements in retrieval tasks on a large-scale e-commerce dataset. We also provide an analysis of the different metadata.

We propose a novel representation for dense pixel-wise estimation tasks using CNNs that boosts accuracy and reduces training time, by explicitly exploiting joint coarse-and-fine reasoning. The coarse reasoning is performed over a discrete classification space to obtain a general rough solution, while the fine details of the solution are obtained over a continuous regression space. In our approach both components are jointly estimated, which proved to be beneficial for improving estimation accuracy. Additionally, we propose a new network architecture, which combines coarse and fine components by treating the fine estimation as a refinement built on top of the coarse solution, and therefore adding details to the general prediction. We apply our approach to the challenging problem of optical flow estimation and empirically validate it against state-of-the-art CNN-based solutions trained from scratch and tested on large optical flow datasets.

Low textured scenes are well known to be one of the main Achilles heels of geometric computer vision algorithms relying on point correspondences, and in particular for visual SLAM. Yet, there are many environments in which, despite being low textured, one can still reliably estimate line-based geometric primitives, for instance in city and indoor scenes, or in the so-called “Manhattan worlds”, where structured edges are predominant. In this paper we propose a solution to handle these situations. Specifically, we build upon ORB-SLAM, presumably the current state-of-the-art solution both in terms of accuracy as efficiency, and extend its formulation to simultaneously handle both point and line correspondences. We propose a solution that can even work when most of the points are vanished out from the input images, and, interestingly it can be initialized from solely the detection of line correspondences in three consecutive frames. We thoroughly evaluate our approach and the new initialization strategy on the TUM RGB-D benchmark and demonstrate that the use of lines does not only improve the performance of the original ORB-SLAM solution in poorly textured frames, but also systematically improves it in sequence frames combining points and lines, without compromising the efficiency.

Most recent approaches to 3D pose estimation from RGB-D images address the problem in a two-stage pipeline. First, they learn a classifier-typically a random forest-to predict the position of each input pixel on the object surface. These estimates are then used to define an energy function that is minimized w.r.t. the object pose. In this paper, we focus on the first stage of the problem and propose a novel classifier based on a depth-aware Convolutional Neural Network. This classifier is able to learn a scale-adaptive regression model that yields very accurate pixel-level predictions, allowing to finally estimate the pose using a simple RANSAC-based scheme, with no need to optimize complex ad hoc energy functions. Our experiments on publicly available datasets show that our approach achieves remarkable improvements over state-of-the-art methods.

In this paper we are interested in recognizing human actions from sequences of 3D skeleton data. For this purpose we combine a 3D Convolutional Neural Network with body representations based on Euclidean Distance Matrices (EDMs), which have been recently shown to be very effective to capture the geometric structure of the human pose. One inherent limitation of the EDMs, however, is that they are defined up to a permutation of the skeleton joints, i.e., randomly shuffling the ordering of the joints yields many different representations. In oder to address this issue we introduce a novel architecture that simultaneously, and in an end-to-end manner, learns an optimal transformation of the joints, while optimizing the rest of parameters of the convolutional network. The proposed approach achieves state-of-the-art results on 3 benchmarks, including the recent NTU RGB-D dataset, for which we improve on previous LSTM-based methods by more than 10 percentage points, also surpassing other CNN-based methods while using almost 1000 times fewer parameters.

The most standard approach to resolve the inherent ambiguities of the non-rigid structure from motion problem is using low-rank models that approximate deforming shapes by a linear combination of rigid basis. These models are typically global, i.e., each shape basis contributes equally to all points of the surface. While this approach has been shown effective to represent smooth deformations, its performance degrades for surfaces composed of various regions, each following a different deformation rule. Piece-wise methods attempt to capture this type of behavior by locally modeling surface patches, although they subsequently require enforcing global constraints to assemble back the patches. In this paper we propose an approach that combines the best of global and local models: it locally considers low-rank models but, by construction, does not need to impose global constraints to guarantee local patch continuity. We achieve this by a simple expectation maximization strategy that besides learning global shape bases, it locally adapts their contribution to each specific surface region. Furthermore, as a side contribution, in order to split the surface into different local patches, we propose a novel physically-based mesh segmentation approach that obeys an energy criterion. The complete framework is evaluated in both synthetic and real datasets, and shows an improved performance to competing methods.

We propose a new superpixel algorithm based on exploiting the boundary information of an image, as objects in images can generally be described by their boundaries. Our proposed approach initially estimates the boundaries and uses them to place superpixel seeds in the areas in which they are more dense. Afterwards, we minimize an energy function in order to expand the seeds into full superpixels. In addition to standard terms such as color consistency and compactness, we propose using the geodesic distance which concentrates small superpixels in regions of the image with more information, while letting larger superpixels cover more homogeneous regions. By both improving the initialization using the boundaries and coherency of the superpixels with geodesic distances, we are able to maintain the coherency of the image structure with fewer superpixels than other approaches. We show the resulting algorithm to yield smaller Variation of Information metrics in seven different datasets while maintaining Undersegmentation Error values similar to the state-of-the-art methods.

Ariadna Quattoni, Arnau Ramisa, Pranava Swaroop Madhyastha, Edgar Simo-Serra, Francesc Moreno-Noguer. Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2016.

This paper describes an on-line approach for estimating non-rigid shape and camera pose from monocular video sequences. We assume an initial estimate of the shape at rest to be given and represented by a triangulated mesh, which is encoded by a matrix of the distances between every pair of vertexes. By applying spectral analysis on this matrix, we are then able to compute a low-dimensional shape basis, that in contrast to standard approaches, has a very direct physical interpretation and requires a much smaller number of modes to span a large variety of deformations, either for inextensible or extensible configurations. Based on this low-rank model, we then sequentially retrieve both camera motion and non-rigid shape in each image, optimizing the model parameters with bundle adjustment over a sliding window of image frames. Since the number of these parameters is small, specially when considering physical priors, our approach may potentially achieve real-time performance. Experimental results on real videos for different scenarios demonstrate remarkable robustness to artifacts such as missing and noisy observations.

We propose a new approach to simultaneously recover camera pose and 3D shape of non-rigid and potentially extensible surfaces from a monocular image sequence. For this purpose, we make use of the Extended Kalman Filter based Simultaneous Localization And Mapping (EKF-SLAM) formulation, a Bayesian optimization framework traditionally used in mobile robotics for estimating camera pose and reconstructing rigid scenarios. In order to extend the problem to a deformable domain we represent the object's surface mechanics by means of Navier's equations, which are solved using a Finite Element Method (FEM). With these main ingredients, we can further model the material's stretching, allowing us to go a step further than most of current techniques, typically constrained to surfaces undergoing isometric deformations. We extensively validate our approach in both real and synthetic experiments, and demonstrate its advantages with respect to competing methods. More specifically, we show that besides simultaneously retrieving camera pose and non-rigid shape, our approach is adequate for both isometric and extensible surfaces, does not require neither batch processing all the frames nor tracking points over the whole sequence and runs at several frames per second.

While recent approaches have shown that it is possible to do template matching by exhaustively scanning the parameter space, the resulting algorithms are still quite demanding. In this paper we alleviate the computational load of these algorithms by proposing an efficient approach for predicting the match ability of a template, before it is actually performed. This avoids large amounts of unnecessary computations. We learn the match ability of templates by using dense convolutional neural network descriptors that do not require ad-hoc criteria to characterize a template. By using deep learning descriptions of patches we are able to predict match ability over the whole image quite reliably. We will also show how no specific training data is required to solve problems like panorama stitching in which you usually require data from the scene in question. Due to the highly parallelizable nature of this tasks we offer an efficient technique with a negligible computational cost at test time.

In this paper, we propose a sequential solution to simultaneously estimate camera pose and non-rigid 3D shape from a monocular video. In contrast to most existing approaches that rely on global representations of the shape, we model the object at a local level, as an ensemble of particles, each ruled by the linear equation of the Newton's second law of motion. This dynamic model is incorporated into a bundle adjustment framework, in combination with simple regularization components that ensure temporal and spatial consistency of the estimated shape and camera poses. The resulting approach is both efficient and robust to several artifacts such as noisy and missing data or sudden camera motions, while it does not require any training data at all. Validation is done in a variety of real video sequences, including articulated and non-rigid motion, both for continuous and discontinuous shapes. Our system is shown to perform comparable to competing batch, computationally expensive, methods and shows remarkable improvement with respect to the sequential ones.

In this paper, we analyze the fashion of clothing of a large social website. Our goal is to learn and predict how fashionable a person looks on a photograph and suggest subtle improvements the user could make to improve her/his appeal. We propose a Conditional Random Field model that jointly reasons about several fashionability factors such as the type of outfit and garments the user is wearing, the type of the user, the photograph's setting (e.g., the scenery behind the user), and the fashionability score. Importantly, our model is able to give rich feedback back to the user, conveying which garments or even scenery she/he should change in order to improve fashionability. We demonstrate that our joint approach significantly outperforms a variety of intelligent baselines. We additionally collected a novel heterogeneous dataset with 144,169 user posts containing diverse image, textual and meta information which can be exploited for our task. We also provide a detailed analysis of the data, showing different outfit trends and fashionability scores across the globe and across a span of 6 years.

We investigate the role that geometric, textual and visual features play in the task of predicting a preposition that links two visual entities depicted in an image.The task is an important part of the subsequent process of generating image descriptions.We explore the prediction of prepositions for a pair of entities, both in the case when the labels of such entities are known and unknown.In all situations we found clear evidence that all three features contribute to the prediction task.

In this paper, we address the problem of simultaneously recovering the 3D shape and pose of a deformable and potentially elastic object from 2D motion. This is a highly ambiguous problem typically tackled by using low-rank shape and trajectory constraints. We show that formulating the problem in terms of a low-rank force space that induces the deformation, allows for a better physical interpretation of the resulting priors and a more accurate representation of the actual object's behavior. However, this comes at the price of, besides force and pose, having to estimate the elastic model of the object. For this, we use an Expectation Maximization strategy, where each of these parameters are successively learned within partial M-steps, while robustly dealing with missing observations. We thoroughly validate the approach on both mocap and real sequences, showing more accurate 3D reconstructions than state-of-the-art, and additionally providing an estimate of the full elastic model with no a priori information.

Deep learning has revolutionalized image-level tasks such as classification, but patch-level tasks, such as correspondence, still rely on hand-crafted features, e.g. SIFT. In this paper we use Convolutional Neural Networks (CNNs) to learn discriminant patch representations and in particular train a Siamese network with pairs of (non-)corresponding patches. We deal with the large number of potential pairs with the combination of a stochastic sampling of the training set and an aggressive mining strategy biased towards patches that are hard to classify. By using the L2 distance during both training and testing we develop 128-D descriptors whose euclidean distances reflect patch similarity, and which can be used as a drop-in replacement for any task involving SIFT. We demonstrate consistent performance gains over the state of the art, and generalize well against scaling and rotation, perspective transformation, non-rigid deformation, and illumination changes. Our descriptors are efficient to compute and amenable to modern GPUs, and are publicly available.

We propose a robust and efficient method to estimate the pose of a camera with respect to complex 3D textured models of the environment that can potentially contain more than 100; 000 points. To tackle this problem we follow a top down approach where we combine high-level deep network classifiers with low level geometric approaches to come up with a solution that is fast, robust and accurate. Given an input image, we initially use a pre-trained deep network to compute a rough estimation of the camera pose. This initial estimate constrains the number of 3D model points that can be seen from the camera viewpoint. We then establish 3D-to-2D correspondences between these potentially visible points of the model and the 2D detected image features. Accurate pose estimation is finally obtained from the 2D-to-3D correspondences using a novel PnP algorithm that rejects outliers without the need to use a RANSAC strategy, and which is between 10 and 100 times faster than other methods that use it. Two real experiments dealing with very large and complex 3D models demonstrate the effectiveness of the approach.

We propose an efficient Human Robot Interaction approach to efficiently model the appearance of all relevant objects in robot's environment. Given an input video stream recorded while the robot is navigating, the user just needs to annotate a very small number of frames to build specific classifiers for each of the objects of interest. At the core of the method, there are several random ferns classifiers that share the same features and are updated online. The resulting methodology is fast (runs at 8 fps), versatile (it can be applied to unconstrained scenarios), scalable (real experiments show we can model up to 30 different object classes), and minimizes the amount of human intervention by leveraging the uncertainty measures associated to each classifier. We thoroughly validate the approach on synthetic data and on real sequences acquired with a mobile platform in outdoor and challenging scenarios containing a multitude of different objects. We show that the human can, with minimal effort, provide the robot with a detailed model of the objects in the scene.

We present a new approach for matching sets of branching curvilinear structures that form graphs embedded in R2 or R3 and may be subject to deformations. Unlike earlier methods, ours does not rely on local appearance similarity nor does require a good initial alignment. Furthermore, it can cope with non-linear deformations, topological differences, and partial graphs. To handle arbitrary non-linear deformations, we use Gaussian process regressions to represent the geometrical mapping relating the two graphs. In the absence of appearance information, we iteratively establish correspondences between points, update the mapping accordingly, and use it to estimate where to find the most likely correspondences that will be used in the next step. To make the computation tractable for large graphs, the set of new potential matches considered at each iteration is not selected at random as with many RANSAC-based algorithms. Instead, we introduce a so-called Active Testing Search strategy that performs a priority search to favor the most likely matches and speed-up the process. We demonstrate the effectiveness of our approach first on synthetic cases and then on angiography data, retinal fundus images, and microscopy image stacks acquired at very different resolutions.

We introduce LETHA (Learning on Easy data, Test on Hard), a new learning paradigm consisting of building strong priors from high quality training data, and combining them with discriminative machine learning to deal with low-quality test data. Our main contribution is an implementation of that concept for pose estimation. We first automatically build a 3D model of the object of interest from high-definition images, and devise from it a pose-indexed feature extraction scheme. We then train a single classifier to process these feature vectors. Given a low quality test image, we visit many hypothetical poses, extract features consistently and evaluate the response of the classifier. Since this process uses locations recorded during learning, it does not require matching points anymore. We use a boosting procedure to train this classifier common to all poses, which is able to deal with missing features, due in this context to self-occlusion. Our results demonstrate that the method combines the strengths of global image representations, discriminative even for very tiny images, and the robustness to occlusions of approaches based on local feature point descriptors.

We propose a real-time, robust to outliers and accurate solution to the Perspective-n-Point (PnP) problem. The main advantages of our solution are twofold: first, it in- tegrates the outlier rejection within the pose estimation pipeline with a negligible computational overhead, and sec- ond, its scalability to arbitrarily large number of correspon- dences. Given a set of 3D-to-2D matches, we formulate pose estimation problem as a low-rank homogeneous sys- tem where the solution lies on its 1D null space. Outlier correspondences are those rows of the linear system which perturb the null space and are progressively detected by projecting them on an iteratively estimated solution of the null space. Since our outlier removal process is based on an algebraic criterion which does not require computing the full-pose and reprojecting back all 3D points on the image plane at each step, we achieve speed gains of more than 100× compared to RANSAC strategies. An extensive exper- imental evaluation will show that our solution yields accu- rate results in situations with up to 50% of outliers, and can process more than 1000 correspondences in less than 5ms.

In this work we propose a technique to combine bottom-up segmentation, coming in the form of SLIC superpixels, with sliding window detectors, such as Deformable Part Models (DPMs). The merit of our approach lies in "cleaning up" the low-level HOG features by exploiting the spatial support of SLIC superpixels, this can be understood as using segmentation to split the feature variation into object-specific and background changes. Rather than committing to a single segmentation we use a large pool of SLIC superpixels and combine them in a scale-, position- and object-dependent manner to build soft segmentation masks. The segmentation masks can be computed fast enough to repeat this process over every candidate window, during training and detection, for both the root and part filters of DPMs. We use these masks to construct enhanced, background-invariant features to train DPMs. We test our approach on the PASCAL VOC 2007, outperforming the standard DPM in 17 out of 20 classes, yielding an average increase of 1.7% AP. Additionally, we demonstrate the robustness of this approach, extending it to dense SIFT descriptors for large displacement optical flow.

We present a real time algorithm for estimating the pose of non-planar objects on which we have placed a visual marker. It is designed to overcome the limitations of small aerial robots, such as slow CPUs, low image resolution and geometric distortions produced by wide angle lenses or viewpoint changes. The method initially registers the shape of a known marker to the contours extracted in an image. For this purpose, and in contrast to state-of-the art, we do not seek to match textured patches or points of interest. Instead, we optimize a geometric alignment cost computed directly from raw polygonal representations of the observed regions using very simple and efficient clipping algorithms. Further speed is achieved by performing the optimization in the polygon representation space, avoiding the need of 2D image processing operations. Deformation modes are easily included in the optimization scheme, allowing an accurate registration of different markers attached to curved surfaces using a single deformable prototype. Once this initial registration is solved, the object pose is retrieved using a standard PnP approach. As a result, the method achieves accurate object pose estimation in real-time, which is very important for interactive UAV tasks, for example for short distance surveillance or bar assembly. We present experiments where our method yields, at about 30Hz, an average error of less than 5mm in estimating the position of a 19×19mm marker placed at 0.7m of the camera.

We present a method for efficiently detecting natural landmarks that can handle scenes with highly repetitive patterns and targets progressively changing its appearance. At the core of our approach lies a Random Ferns classifier, that models the posterior probabilities of different views of the target using multiple and independent Ferns, each containing features at particular positions of the target. A Shannon entropy measure is used to pick the most informative locations of these features. This minimizes the number of Ferns while maximizing its discriminative power, allowing thus, for robust detections at low computational costs. In addition, after offline initialization, the new incoming detections are used to update the posterior probabilities on the fly, and adapt to changing appearances that can occur due to the presence of shadows or occluding objects. All these virtues, make the proposed detector appropriate for UAV navigation. Besides the synthetic experiments that will demonstrate the theoretical benefits of our formulation, we will show applications for detecting landing areas in regions with highly repetitive patterns, and specific objects under the presence of cast shadows or sudden camera motions.

Presentado al 24th BMVC celebrado en Bristol (UK) del 9 al 13 de septiembre 2013.-- The copyright of this document resides with its authors.

We introduce a novel approach to automatically recover 3D human pose from a single image. Most previous work follows a pipelined approach: initially, a set of 2D features such as edges, joints or silhouettes are detected in the image, and then these observations are used to infer the 3D pose. Solving these two problems separately may lead to erroneous 3D poses when the feature detector has performed poorly. In this paper, we address this issue by jointly solving both the 2D detection and the 3D inference problems. For this purpose, we propose a Bayesian framework that integrates a generative model based on latent variables and discriminative 2D part detectors based on HOGs, and perform inference using evolutionary algorithms. Real experimentation demonstrates competitive results, and the ability of our methodology to provide accurate 2D and 3D pose estimations even when the 2D detectors are inaccurate.

In this work we exploit segmentation to construct appearance descriptors that can robustly deal with occlusion and background changes. For this, we downplay measurements coming from areas that are unlikely to belong to the same region as the descriptor's center, as suggested by soft segmentation masks. Our treatment is applicable to any image point, i.e. dense, and its computational overhead is in the order of a few seconds. We integrate this idea with Dense SIFT, and also with Dense Scale and Rotation Invariant Descriptors (SID), delivering descriptors that are densely computable, invariant to scaling and rotation, and robust to background changes. We apply our approach to standard benchmarks on large displacement motion estimation using SIFT-flow and wide-baseline stereo, systematically demonstrating that the introduction of segmentation yields clear improvements.

Most current depth sensors provide 2.5D range images in which depth values are assigned to a rectangular 2D array. In this paper we take advantage of this structured information to build an efficient shape descriptor which is about two orders of magnitude faster than competing approaches, while showing similar performance in several tasks involving deformable object recognition. Given a 2D patch surrounding a point and its associated depth values, we build the descriptor for that point, based on the cumulative distances between their normals and a discrete set of normal directions. This processing is made very efficient using integral images, even allowing to compute descriptors for every range image pixel in a few seconds. The discriminative power of our descriptor, dubbed FINDDD, is evaluated in three different scenarios: recognition of specific cloth wrinkles, instance recognition from geometry alone, and detection of reliable and informed grasping points.

During the last decade, there has been a growing interest in making autonomous social robots able to interact with people. However, there are still many open issues regarding the social capabilities that robots should have in order to perform these interactions more naturally. In this paper we present the results of several experiments conducted at the Barcelona Robot Lab in the campus of the “Universitat Politècnica de Catalunya” in which we have analyzed different important aspects of the interaction between a mobile robot and nontrained human volunteers. First, we have proposed different robot behaviors to approach a person and create an engagement with him/her. In order to perform this task we have provided the robot with several perception and action capabilities, such as that of detecting people, planning an approach and verbally communicating its intention to initiate a conversation. Once the initial engagement has been created, we have developed further communication skills in order to let people assist the robot and improve its face recognition system. After this assisted and online learning stage, the robot becomes able to detect people under severe changing conditions, which, in turn enhances the number and the manner that subsequent human-robot interactions are performed.

Recovering the 3D shape of deformable surfaces from single images is known to be a highly ambiguous problem because many different shapes may have very similar projections. This is commonly addressed by restricting the set of possible shapes to linear combinations of deformation modes and by imposing additional geometric constraints. Unfortunately, because image measurements are noisy, such constraints do not always guarantee that the correct shape will be recovered. To overcome this limitation, we introduce a stochastic sampling approach to efficiently explore the set of solutions of an objective function based on point correspondences. This allows us to propose a small set of ambiguous candidate 3D shapes and then use additional image information to choose the best one. As a proof of concept, we use either motion or shading cues to this end and show that we can handle a complex objective function without having to solve a difficult nonlinear minimization problem. The advantages of our method are demonstrated on a variety of problems including both real and synthetic data.

We propose a novel approach for the estimation of the pose and focal length of a camera from a set of 3D-to-2D point correspondences. Our method compares favorably to competing approaches in that it is both more accurate than existing closed form solutions, as well as faster and also more accurate than iterative ones. Our approach is inspired on the EPnP algorithm, a recent O(n) solution for the calibrated case. Yet we show that considering the focal length as an additional unknown renders the linearization and relinearization techniques of the original approach no longer valid, especially with large amounts of noise. We present new methodologies to circumvent this limitation termed exhaustive linearization and exhaustive relinearization which perform a systematic exploration of the solution space in closed form. The method is evaluated on both real and synthetic data, and our results show that besides producing precise focal length estimation, the retrieved camera pose is almost as accurate as the one computed using the EPnP, which assumes a calibrated camera.

We present a new approach to matching graphs embedded in ℝ2or ℝ3. Unlike earlier methods, our approach does not rely on the similarity of local appearance features, does not require an initial alignment, can handle partial matches, and can cope with non-linear deformations and topological differences. To handle arbitrary non-linear deformations, we represent them as Gaussian Processes. In the absence of appearance information, we iteratively establish correspondences between graph nodes, update the structure accordingly, and use the current mapping estimate to find the most likely correspondences that will be used in the next iteration. This makes the computation tractable. We demonstrate the effectiveness of our approach first on synthetic cases and then on angiography data, retinal fundus images, and microscopy image stacks acquired at very different resolutions.

Markerless 3D human pose detection from a single image is a severely underconstrained problem because different 3D poses can have similar image projections. In order to handle this ambiguity, current approaches rely on prior shape models that can only be correctly adjusted if 2D image features are accurately detected. Unfortunately, although current 2D part detector algorithms have shown promising results, they are not yet accurate enough to guarantee a complete disambiguation of the 3D inferred shape. In this paper, we introduce a novel approach for estimating 3D human pose even when observations are noisy. We propose a stochastic sampling strategy to propagate the noise from the image plane to the shape space. This provides a set of ambiguous 3D shapes, which are virtually undistinguishable from their image projections. Disambiguation is then achieved by imposing kinematic constraints that guarantee the resulting pose resembles a 3D human shape. We validate the method on a variety of situations in which state-of-the-art 2D detectors yield either inaccurate estimations or partly miss some of the body parts.

Detecting grasping points is a key problem in cloth manipulation. Most current approaches follow a multiple re-grasp strategy for this purpose, in which clothes are sequentially grasped from different points until one of them yields to a desired configuration. In this paper, by contrast, we circumvent the need for multiple re-graspings by building a robust detector that identifies the grasping points, generally in one single step, even when clothes are highly wrinkled. In order to handle the large variability a deformed cloth may have, we build a Bag of Features based detector that combines appearance and 3D geometry features. An image is scanned using a sliding window with a linear classifier, and the candidate windows are refined using a non-linear SVM and a “grasp goodness” criterion to select the best grasping point. We demonstrate our approach detecting collars in deformed polo shirts, using a Kinect camera. Experimental results show a good performance of the proposed method not only in identifying the same trained textile object part under severe deformations and occlusions, but also the corresponding part in other clothes, exhibiting a degree of generalization.

We propose an efficient method for object localization and 3D pose estimation. A two-step approach is used. In the first step, a pose estimator is evaluated in the input images in order to estimate potential object locations and poses. These candidates are then validated, in the second step, by the corresponding pose-specific classifier. The result is a detection approach that avoids the inherent and expensive cost of testing the complete set of specific classifiers over the entire image. A further speedup is achieved by feature sharing. Features are computed only once and are then used for evaluating the pose estimator and all specific classifiers. The proposed method has been validated on two public datasets for the problem of detecting of cars under several views. The results show that the proposed approach yields high detection rates while keeping efficiency. © 2011. The copyright of this document resides with its authors.

Recent advances in 3D shape recognition have shown that kernels based on diffusion geometry can be effectively used to describe local features of deforming surfaces. In this paper, we introduce a new framework that allows using these kernels on 2D local patches, yielding a novel feature point descriptor that is both invariant to non-rigid image deformations and illumination changes. In order to build the descriptor, 2D image patches are embedded as 3D surfaces, by multiplying the intensity level by an arbitrarily large and constant weight that favors anisotropic diffusion and retains the gradient magnitude information. Patches are then described in terms of a heat kernel signature, which is made invariant to intensity changes, rotation and scaling. The resulting feature point descriptor is proven to be significantly more discriminative than state of the art ones, even those which are specifically designed for describing non-rigid image deformations.

We present an algorithm to simultaneously recover non-rigid shape and camera poses from point correspondences between a reference shape and a sequence of input images. The key novel contribution of our approach is in bringing the tools of the probabilistic SLAM methodology from a rigid to a deformable domain. Under the assumption that the shape may be represented as a weighted sum of deformation modes, we show that the problem of estimating the modal weights along with the camera poses, may be probabilistically formulated as a maximum a posterior estimate and solved using an iterative least squares optimization. An extensive evaluation on synthetic and real data, shows that our approach has several significant advantages over current approaches, such as performing robustly under large amounts of noise and outliers, and neither requiring to track points over the whole sequence nor initializations close from the ground truth solution.

We present a novel approach to simultaneously reconstruct the 3D structure of a non-rigid coronary tree and estimate point correspondences between an input X-ray image and a reference 3D shape. At the core of our approach lies an optimization scheme that iteratively fits a generative 3D model of increasing complexity and guides the matching process. As a result, and in contrast to existing approaches that assume rigidity or quasi-rigidity of the structure, our method is able to retrieve large non-linear deformations even when the input data is corrupted by the presence of noise and partial occlusions. We extensively evaluate our approach under synthetic and real data and demonstrate a remarkable improvement compared to state-of-the-art.

Recent works have shown that 3D shape of non-rigid surfaces can be accurately retrieved from a single image given a set of 3D-to-2D correspondences between that image and another one for which the shape is known. However, existing approaches assume that such correspondences can be readily established, which is not necessarily true when large deformations produce significant appearance changes between the input and the reference images. Furthermore, it is either assumed that the pose of the camera is known, or the estimated solution is pose-ambiguous. In this paper we relax all these assumptions and, given a set of 3D and 2D unmatched points, we present an approach to simultaneously solve their correspondences, compute the camera pose and retrieve the shape of the surface in the input image. This is achieved by introducing weak priors on the pose and shape that we model as Gaussian Mixtures. By combining them into a Kalman filter we can progressively reduce the number of 2D candidates that can be potentially matched to each 3D point, while pose and shape are refined. This lets us to perform a complete and efficient exploration of the solution space and retain the best solution.

We present a new approach for building an efficient and robust classifier for the two class problem, that localizes objects that may appear in the image under different orientations. In contrast to other works that address this problem using multiple classifiers, each one specialized for a specific orientation, we propose a simple two-step approach with an estimation stage and a classification stage. The estimator yields an initial set of potential object poses that are then validated by the classifier. This methodology allows reducing the time complexity of the algorithm while classification results remain high. The classifier we use in both stages is based on a boosted combination of Random Ferns over local histograms of oriented gradients (HOGs), which we compute during a preprocessing step. Both the use of supervised learning and working on the gradient space makes our approach robust while being efficient at run-time. We show these properties by thorough testing on standard databases and on a new database made of motorbikes under planar rotations, and with challenging conditions such as cluttered backgrounds, changing illumination conditions and partial occlusions.

We propose a new algorithm for detecting multiple object categories that exploits the fact that different categories may share common features but with different geometric distributions. This yields an efficient detector which, in contrast to existing approaches, considerably reduces the computation cost at runtime, where the feature computation step is traditionally the most expensive. More specifically, at the learning stage we compute common features by applying the same Random Ferns over the Histograms of Oriented Gradients on the training images. We then apply a boosting step to build discriminative weak classifiers, and learn the specific geometric distribution of the Random Ferns for each class. At runtime, only a few Random Ferns have to be densely computed over each input image, and their geometric distribution allows performing the detection. The proposed method has been validated in public datasets achieving competitive detection results, which are comparable with state-of-the-art methods that use specific features per class.

We present a closed form solution to the problem of recovering the 3D shape of a nonrigid potentially stretchable surface from 3D-to-2D correspondences. In other words, we can reconstruct a surface from a single image without a priori knowledge of its deformations in that image. State of the art solutions to nonrigid 3D shape recovery rely on the fact that distances between neighboring surface points must be preserved and are therefore limited to inelastic surfaces. Here, we show that replacing the inextensibility constraints by shading ones removes this limitation while still allowing 3D reconstruction in closed-form. We demonstrate our method and compare it to an earlier one using both synthetic and real data.

We present a new model for people guidance in urban settings using several mobile robots, that overcomes the limitations of existing approaches, which are either tailored to tightly bounded environments, or based on unrealistic human behaviors. Although the robots motion is controlled by means of a standard particle filter formulation, the novelty of our approach resides in how the environment and human and robot motions are modeled. In particular we define a ¿Discrete-Time- Motion¿ model, which from one side represents the environment by means of a potential field, that makes it appropriate to deal with open areas, and on the other hand the motion models for people and robots respond to realistic situations, and for instance human behaviors such as ¿leaving the group¿ are considered.

We propose a new technique for fusing multiple cues to robustly segment an object from its background in video sequences that suffer from abrupt changes of both illumination and position of the target. Robustness is achieved by the integration of appearance and geometric object features and by their estimation using Bayesian filters, such as Kalman or particle filters. In particular, each filter estimates the state of a specific object feature, conditionally dependent on another feature estimated by a distinct filter. This dependence provides improved target representations, permitting to segment it out from the background even in non-stationary sequences. Considering that the procedure of the Bayesian filters may be described by a "hypotheses generation--hypotheses correction" strategy, the major novelty of our methodology compared to previous approaches is that the mutual dependence between filters is considered during the feature observation, i.e, into the "hypotheses correction" stage,instead of considering it when generating the hypotheses. This proves to be much more effective in terms of accuracy and reliability. The proposed method is analytically justified and applied to develop a robust tracking system that adapts online and simultaneously the color space where the image points are represented, the color distributions, the contour of the object and its bounding box. Results with synthetic data and real video sequences demonstrate the robustness and versatility of our method.

We propose a non-iterative solution to the PnP problem-the estimation of the pose of a calibrated camera from n 3D-to-2D point correspondences—whose computational complexity grows linearly with 𝑛5) or even 𝑂(𝑛8), without being more accurate. Our method is applicable for all 𝑛≥4 and handles properly both planar and non-planar configurations. Our central idea is to express the 𝑛 3D points as a weighted sum of four virtual control points. The problem then reduces to estimating the coordinates of these control points in the camera referential, which can be done in 𝑂(𝑛) time by expressing these coordinates as weighted sum of the eigenvectors of a 12 × 12 matrix and solving a small constant number of quadratic equations to pick the right weights. The advantages of our method are demonstrated by thorough testing on both synthetic and real-data.

We present a system for refocusing images and videos of dynamic scenes using a novel, single-view depth estimation method. Our method for obtaining depth is based on the defocus of a sparse set of dots projected onto the scene. In contrast to other active illumination techniques, the projected pattern of dots can be removed from each captured image and its brightness easily controlled in order to avoid under- or over-exposure. The depths corresponding to the projected dots and a color segmentation of the image are used to compute an approximate depth map of the scene with clean region boundaries. The depth map is used to refocus the acquired image after the dots are removed, simulating realistic depth of field effects. Experiments on a wide variety of scenes, including close-ups and live action, demonstrate the effectiveness of our method.

Robotics applications based on computer vision algorithms are highly constrained to indoor environments where conditions may be controlled. The development of robust visual algorithms is necessary for improving the capabilities of many autonomous systems in outdoor and dynamic environments. In particular, this paper proposes a tracking algorithm robust to several artifacts which may be found in real world applications, such as lighting changes, cluttered backgrounds and unexpected target movements. In order to deal with these difficulties the proposed tracking methodology integrates several Bayesian filters. Each filter estimates the state of a particular object feature which is conditionally dependent on another feature estimated by a distinct filter. This dependence provides improved representations of the target, allowing to segment it out from the background of the image. We describe the updating procedure of the Bayesian filters by a 'hypotheses generation and correction' scheme. The main difference with respect to previous approaches is that the dependence between filters is considered during the feature observation, i.e., into the 'hypotheses correction' stage, instead of considering it when generating the hypotheses. This proves to be much more effective in terms of accuracy and reliability

We propose a new technique for focusing multiple cues to robustly segment an object from its background in video sequences that suffer from abrupt changes of both illumination and position of the target. Robustness is achieved by tile integration of appearance and geometric object features and by their description using particle filters. Previous approaches assume independence of the object cues or apply the particle filter formulation to only one of the features, and assume a smooth change in the rest, which can prove is very limiting, especially when the state of some features needs to be updated using other cues or when their dynamics follow non-linear and unpredictable paths. Our technique offers a general framework to model the probabilistic relationship between features. The proposed method is analytically justified and applied to develop a robust tracking system that adapts online and simultaneously the color space where the image points are represented, the color distributions, and the contour of the object. Results with synthetic data and real video sequences demonstrate the robustness and versatility of our method

Describes a method to localize faces in color images based on the fusion of the information gathered from a stereo vision system and the analysis of color images. Our method generates a depth map of the scene and tries to fit a head model taking into account the shape of the model and skin color information. The method is tailored for its use in factory automation applications where the detection and localization of humans is necessary for the completion or interruption of a particular task, such as robot manipulator safety or the interaction of service robots with humans.