Philippe Weinzaepfel

Martin Humenberger, Yohann Cabon, Noé Pion et al.

Visual localization, i.e., camera pose estimation in a known scene, is a core component of technologies such as autonomous driving and augmented reality. State-of-the-art localization approaches often rely on image retrieval techniques for one of two purposes: (1) provide an approximate pose estimate or (2) determine which parts of the scene are potentially visible in a given query image. It is common practice to use state-of-the-art image retrieval algorithms for both of them. These algorithms are often trained for the goal of retrieving the same landmark under a large range of viewpoint changes which often differs from the requirements of visual localization. In order to investigate the consequences for visual localization, this paper focuses on understanding the role of image retrieval for multiple visual localization paradigms. First, we introduce a novel benchmark setup and compare state-of-the-art retrieval representations on multiple datasets using localization performance as metric. Second, we investigate several definitions of "ground truth" for image retrieval. Using these definitions as upper bounds for the visual localization paradigms, we show that there is still sgnificant room for improvement. Third, using these tools and in-depth analysis, we show that retrieval performance on classical landmark retrieval or place recognition tasks correlates only for some but not all paradigms to localization performance. Finally, we analyze the effects of blur and dynamic scenes in the images. We conclude that there is a need for retrieval approaches specifically designed for localization paradigms. Our benchmark and evaluation protocols are available at https://github.com/naver/kapture-localization.

Fabien Baradel, Romain Brégier, Thibault Groueix et al.

Training state-of-the-art models for human pose estimation in videos requires datasets with annotations that are really hard and expensive to obtain. Although transformers have been recently utilized for body pose sequence modeling, related methods rely on pseudo-ground truth to augment the currently limited training data available for learning such models. In this paper, we introduce PoseBERT, a transformer module that is fully trained on 3D Motion Capture (MoCap) data via masked modeling. It is simple, generic and versatile, as it can be plugged on top of any image-based model to transform it in a video-based model leveraging temporal information. We showcase variants of PoseBERT with different inputs varying from 3D skeleton keypoints to rotations of a 3D parametric model for either the full body (SMPL) or just the hands (MANO). Since PoseBERT training is task agnostic, the model can be applied to several tasks such as pose refinement, future pose prediction or motion completion without finetuning. Our experimental results validate that adding PoseBERT on top of various state-of-the-art pose estimation methods consistently improves their performances, while its low computational cost allows us to use it in a real-time demo for smoothly animating a robotic hand via a webcam. Test code and models are available at https://github.com/naver/posebert.

Philippe Weinzaepfel, Thomas Lucas, Vincent Leroy et al.

Despite impressive performance for high-level downstream tasks, self-supervised pre-training methods have not yet fully delivered on dense geometric vision tasks such as stereo matching or optical flow. The application of self-supervised concepts, such as instance discrimination or masked image modeling, to geometric tasks is an active area of research. In this work, we build on the recent cross-view completion framework, a variation of masked image modeling that leverages a second view from the same scene which makes it well suited for binocular downstream tasks. The applicability of this concept has so far been limited in at least two ways: (a) by the difficulty of collecting real-world image pairs -- in practice only synthetic data have been used -- and (b) by the lack of generalization of vanilla transformers to dense downstream tasks for which relative position is more meaningful than absolute position. We explore three avenues of improvement. First, we introduce a method to collect suitable real-world image pairs at large scale. Second, we experiment with relative positional embeddings and show that they enable vision transformers to perform substantially better. Third, we scale up vision transformer based cross-completion architectures, which is made possible by the use of large amounts of data. With these improvements, we show for the first time that state-of-the-art results on stereo matching and optical flow can be reached without using any classical task-specific techniques like correlation volume, iterative estimation, image warping or multi-scale reasoning, thus paving the way towards universal vision models.

Philippe Weinzaepfel, Vincent Leroy, Thomas Lucas et al.

Masked Image Modeling (MIM) has recently been established as a potent pre-training paradigm. A pretext task is constructed by masking patches in an input image, and this masked content is then predicted by a neural network using visible patches as sole input. This pre-training leads to state-of-the-art performance when finetuned for high-level semantic tasks, e.g. image classification and object detection. In this paper we instead seek to learn representations that transfer well to a wide variety of 3D vision and lower-level geometric downstream tasks, such as depth prediction or optical flow estimation. Inspired by MIM, we propose an unsupervised representation learning task trained from pairs of images showing the same scene from different viewpoints. More precisely, we propose the pretext task of cross-view completion where the first input image is partially masked, and this masked content has to be reconstructed from the visible content and the second image. In single-view MIM, the masked content often cannot be inferred precisely from the visible portion only, so the model learns to act as a prior influenced by high-level semantics. In contrast, this ambiguity can be resolved with cross-view completion from the second unmasked image, on the condition that the model is able to understand the spatial relationship between the two images. Our experiments show that our pretext task leads to significantly improved performance for monocular 3D vision downstream tasks such as depth estimation. In addition, our model can be directly applied to binocular downstream tasks like optical flow or relative camera pose estimation, for which we obtain competitive results without bells and whistles, i.e., using a generic architecture without any task-specific design.

Bardienus Duisterhof, Lojze Zust, Philippe Weinzaepfel et al.

Structure-from-Motion (SfM), a task aiming at jointly recovering camera poses and 3D geometry of a scene given a set of images, remains a hard problem with still many open challenges despite decades of significant progress. The traditional solution for SfM consists of a complex pipeline of minimal solvers which tends to propagate errors and fails when images do not sufficiently overlap, have too little motion, etc. Recent methods have attempted to revisit this paradigm, but we empirically show that they fall short of fixing these core issues. In this paper, we propose instead to build upon a recently released foundation model for 3D vision that can robustly produce local 3D reconstructions and accurate matches. We introduce a low-memory approach to accurately align these local reconstructions in a global coordinate system. We further show that such foundation models can serve as efficient image retrievers without any overhead, reducing the overall complexity from quadratic to linear. Overall, our novel SfM pipeline is simple, scalable, fast and truly unconstrained, i.e. it can handle any collection of images, ordered or not. Extensive experiments on multiple benchmarks show that our method provides steady performance across diverse settings, especially outperforming existing methods in small- and medium-scale settings.

Thomas Lucas, Fabien Baradel, Philippe Weinzaepfel et al.

We address the problem of action-conditioned generation of human motion sequences. Existing work falls into two categories: forecast models conditioned on observed past motions, or generative models conditioned on action labels and duration only. In contrast, we generate motion conditioned on observations of arbitrary length, including none. To solve this generalized problem, we propose PoseGPT, an auto-regressive transformer-based approach which internally compresses human motion into quantized latent sequences. An auto-encoder first maps human motion to latent index sequences in a discrete space, and vice-versa. Inspired by the Generative Pretrained Transformer (GPT), we propose to train a GPT-like model for next-index prediction in that space; this allows PoseGPT to output distributions on possible futures, with or without conditioning on past motion. The discrete and compressed nature of the latent space allows the GPT-like model to focus on long-range signal, as it removes low-level redundancy in the input signal. Predicting discrete indices also alleviates the common pitfall of predicting averaged poses, a typical failure case when regressing continuous values, as the average of discrete targets is not a target itself. Our experimental results show that our proposed approach achieves state-of-the-art results on HumanAct12, a standard but small scale dataset, as well as on BABEL, a recent large scale MoCap dataset, and on GRAB, a human-object interactions dataset.

Ginger Delmas, Philippe Weinzaepfel, Francesc Moreno-Noguer et al.

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 PoseFix 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.

Mert Bulent Sariyildiz, Philippe Weinzaepfel, Thomas Lucas et al.

Pretrained models have become a commodity and offer strong results on a broad range of tasks. In this work, we focus on classification and seek to learn a unique encoder able to take from several complementary pretrained models. We aim at even stronger generalization across a variety of classification tasks. We propose to learn such an encoder via multi-teacher distillation. We first thoroughly analyse standard distillation when driven by multiple strong teachers with complementary strengths. Guided by this analysis, we gradually propose improvements to the basic distillation setup. Among those, we enrich the architecture of the encoder with a ladder of expendable projectors, which increases the impact of intermediate features during distillation, and we introduce teacher dropping, a regularization mechanism that better balances the teachers' influence. Our final distillation strategy leads to student models of the same capacity as any of the teachers, while retaining or improving upon the performance of the best teacher for each task. Project page and code: https://europe.naverlabs.com/unic

Jerome Revaud, Yohann Cabon, Romain Brégier et al.

Scene coordinates regression (SCR), i.e., predicting 3D coordinates for every pixel of a given image, has recently shown promising potential. However, existing methods remain limited to small scenes memorized during training, and thus hardly scale to realistic datasets and scenarios. In this paper, we propose a generalized SCR model trained once to be deployed in new test scenes, regardless of their scale, without any finetuning. Instead of encoding the scene coordinates into the network weights, our model takes as input a database image with some sparse 2D pixel to 3D coordinate annotations, extracted from e.g. off-the-shelf Structure-from-Motion or RGB-D data, and a query image for which are predicted a dense 3D coordinate map and its confidence, based on cross-attention. At test time, we rely on existing off-the-shelf image retrieval systems and fuse the predictions from a shortlist of relevant database images w.r.t. the query. Afterwards camera pose is obtained using standard Perspective-n-Point (PnP). Starting from selfsupervised CroCo pretrained weights, we train our model on diverse datasets to ensure generalizabilty across various scenarios, and significantly outperform other scene regression approaches, including scene-specific models, on multiple visual localization benchmarks. Finally, we show that the database representation of images and their 2D-3D annotations can be highly compressed with negligible loss of localization performance.

Guillaume Bono, Leonid Antsfeld, Boris Chidlovskii et al.

Most recent work in goal oriented visual navigation resorts to large-scale machine learning in simulated environments. The main challenge lies in learning compact representations generalizable to unseen environments and in learning high-capacity perception modules capable of reasoning on high-dimensional input. The latter is particularly difficult when the goal is not given as a category ("ObjectNav") but as an exemplar image ("ImageNav"), as the perception module needs to learn a comparison strategy requiring to solve an underlying visual correspondence problem. This has been shown to be difficult from reward alone or with standard auxiliary tasks. We address this problem through a sequence of two pretext tasks, which serve as a prior for what we argue is one of the main bottleneck in perception, extremely wide-baseline relative pose estimation and visibility prediction in complex scenes. The first pretext task, cross-view completion is a proxy for the underlying visual correspondence problem, while the second task addresses goal detection and finding directly. We propose a new dual encoder with a large-capacity binocular ViT model and show that correspondence solutions naturally emerge from the training signals. Experiments show significant improvements and SOTA performance on the two benchmarks, ImageNav and the Instance-ImageNav variant, where camera intrinsics and height differ between observation and goal.

Anilkumar Swamy, Vincent Leroy, Philippe Weinzaepfel et al.

Recent hand-object interaction datasets show limited real object variability and rely on fitting the MANO parametric model to obtain groundtruth hand shapes. To go beyond these limitations and spur further research, we introduce the SHOWMe dataset which consists of 96 videos, annotated with real and detailed hand-object 3D textured meshes. Following recent work, we consider a rigid hand-object scenario, in which the pose of the hand with respect to the object remains constant during the whole video sequence. This assumption allows us to register sub-millimetre-precise groundtruth 3D scans to the image sequences in SHOWMe. Although simpler, this hypothesis makes sense in terms of applications where the required accuracy and level of detail is important eg., object hand-over in human-robot collaboration, object scanning, or manipulation and contact point analysis. Importantly, the rigidity of the hand-object systems allows to tackle video-based 3D reconstruction of unknown hand-held objects using a 2-stage pipeline consisting of a rigid registration step followed by a multi-view reconstruction (MVR) part. We carefully evaluate a set of non-trivial baselines for these two stages and show that it is possible to achieve promising object-agnostic 3D hand-object reconstructions employing an SfM toolbox or a hand pose estimator to recover the rigid transforms and off-the-shelf MVR algorithms. However, these methods remain sensitive to the initial camera pose estimates which might be imprecise due to lack of textures on the objects or heavy occlusions of the hands, leaving room for improvements in the reconstruction. Code and dataset are available at https://europe.naverlabs.com/research/showme

Ginger Delmas, Philippe Weinzaepfel, Thomas Lucas et al.

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.

Ginger Delmas, Philippe Weinzaepfel, Francesc Moreno-Noguer et al.

Aligning multiple modalities in a latent space, such as images and texts, has shown to produce powerful semantic visual representations, fueling tasks like image captioning, text-to-image generation, or image grounding. In the context of human-centric vision, albeit CLIP-like representations encode most standard human poses relatively well (such as standing or sitting), they lack sufficient acuteness to discern detailed or uncommon ones. Actually, while 3D human poses have been often associated with images (e.g. to perform pose estimation or pose-conditioned image generation), or more recently with text (e.g. for text-to-pose generation), they have seldom been paired with both. In this work, we combine 3D poses, person's pictures and textual pose descriptions to produce an enhanced 3D-, visual- and semantic-aware human pose representation. We introduce a new transformer-based model, trained in a retrieval fashion, which can take as input any combination of the aforementioned modalities. When composing modalities, it outperforms a standard multi-modal alignment retrieval model, making it possible to sort out partial information (e.g. image with the lower body occluded). We showcase the potential of such an embroidered pose representation for (1) SMPL regression from image with optional text cue; and (2) on the task of fine-grained instruction generation, which consists in generating a text that describes how to move from one 3D pose to another (as a fitness coach). Unlike prior works, our model can take any kind of input (image and/or pose) without retraining.

Matthieu Armando, Salma Galaaoui, Fabien Baradel et al.

Human perception and understanding is a major domain of computer vision which, like many other vision subdomains recently, stands to gain from the use of large models pre-trained on large datasets. We hypothesize that the most common pre-training strategy of relying on general purpose, object-centric image datasets such as ImageNet, is limited by an important domain shift. On the other hand, collecting domain-specific ground truth such as 2D or 3D labels does not scale well. Therefore, we propose a pre-training approach based on self-supervised learning that works on human-centric data using only images. Our method uses pairs of images of humans: the first is partially masked and the model is trained to reconstruct the masked parts given the visible ones and a second image. It relies on both stereoscopic (cross-view) pairs, and temporal (cross-pose) pairs taken from videos, in order to learn priors about 3D as well as human motion. We pre-train a model for body-centric tasks and one for hand-centric tasks. With a generic transformer architecture, these models outperform existing self-supervised pre-training methods on a wide set of human-centric downstream tasks, and obtain state-of-the-art performance for instance when fine-tuning for model-based and model-free human mesh recovery.

Vincent Leroy, Jerome Revaud, Thomas Lucas et al.

Transformers have become the standard in state-of-the-art vision architectures, achieving impressive performance on both image-level and dense pixelwise tasks. However, training vision transformers for high-resolution pixelwise tasks has a prohibitive cost. Typical solutions boil down to hierarchical architectures, fast and approximate attention, or training on low-resolution crops. This latter solution does not constrain architectural choices, but it leads to a clear performance drop when testing at resolutions significantly higher than that used for training, thus requiring ad-hoc and slow post-processing schemes. In this paper, we propose a novel strategy for efficient training and inference of high-resolution vision transformers. The key principle is to mask out most of the high-resolution inputs during training, keeping only N random windows. This allows the model to learn local interactions between tokens inside each window, and global interactions between tokens from different windows. As a result, the model can directly process the high-resolution input at test time without any special trick. We show that this strategy is effective when using relative positional embedding such as rotary embeddings. It is 4 times faster to train than a full-resolution network, and it is straightforward to use at test time compared to existing approaches. We apply this strategy to three dense prediction tasks with high-resolution data. First, we show on the task of semantic segmentation that a simple setting with 2 windows performs best, hence the name of our method: Win-Win. Second, we confirm this result on the task of monocular depth prediction. Third, we further extend it to the binocular task of optical flow, reaching state-of-the-art performance on the Spring benchmark that contains Full-HD images with an order of magnitude faster inference than the best competitor.

Pau de Jorge, César Roberto de Souza, Björn Michele et al.

Efficiently merging several models fine-tuned for different tasks, but stemming from the same pretrained base model, is of great practical interest. Despite extensive prior work, most evaluations of model merging in computer vision are restricted to image classification using CLIP, where different classification datasets define different tasks. In this work, our goal is to make model merging more practical and show its relevance on challenging scenarios beyond this specific setting. In most vision scenarios, different tasks rely on trainable and usually heterogeneous decoders. Differently from previous studies with frozen decoders, where merged models can be evaluated right away, the non-trivial cost of decoder training renders hyperparameter selection based on downstream performance impractical. To address this, we introduce the task alignment proxy, and show how it can be used to speed up hyperparameter selection by orders of magnitude while retaining performance. Equipped with the task alignment proxy, we extend the applicability of model merging to multi-task vision models beyond CLIP-based classification.

Romain Brégier, Guénolé Fiche, Laura Bravo-Sánchez et al.

Parametric body models are central to many human-centric tasks, yet existing models often rely on costly 3D scans and learned shape spaces that are proprietary and demographically narrow. We introduce Anny, a simple, fully differentiable, and scan-free human body model grounded in anthropometric knowledge from the MakeHuman community. Anny defines a continuous, interpretable shape space, where phenotype parameters (e.g. gender, age, height, weight) control blendshapes spanning a wide range of human forms -- across ages (from infants to elders), body types, and proportions. Calibrated using WHO population statistics, it provides realistic and demographically grounded human shape variation within a single unified model. Thanks to its openness and semantic control, Anny serves as a versatile foundation for 3D human modeling -- supporting millimeter-accurate scan fitting, controlled synthetic data generation, and Human Mesh Recovery (HMR). We further introduce Anny-One, a collection of 800k photorealistic humans generated with Anny, showing that despite its simplicity, HMR models trained with Anny can match the performance of those trained with scan-based body models, while remaining interpretable and broadly representative. The Anny body model and its code are released under the Apache 2.0 license, making Anny an accessible foundation for human-centric 3D modeling.

Philippe Weinzaepfel, Thomas Lucas, Diane Larlus et al.

Methods that combine local and global features have recently shown excellent performance on multiple challenging deep image retrieval benchmarks, but their use of local features raises at least two issues. First, these local features simply boil down to the localized map activations of a neural network, and hence can be extremely redundant. Second, they are typically trained with a global loss that only acts on top of an aggregation of local features; by contrast, testing is based on local feature matching, which creates a discrepancy between training and testing. In this paper, we propose a novel architecture for deep image retrieval, based solely on mid-level features that we call Super-features. These Super-features are constructed by an iterative attention module and constitute an ordered set in which each element focuses on a localized and discriminant image pattern. For training, they require only image labels. A contrastive loss operates directly at the level of Super-features and focuses on those that match across images. A second complementary loss encourages diversity. Experiments on common landmark retrieval benchmarks validate that Super-features substantially outperform state-of-the-art methods when using the same number of features, and only require a significantly smaller memory footprint to match their performance. Code and models are available at: https://github.com/naver/FIRe.

Jens Lundell, Enric Corona, Tran Nguyen Le et al.

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/.

Fabien Baradel, Matthieu Armando, Salma Galaaoui et al.

We present Multi-HMR, a strong sigle-shot model for multi-person 3D human mesh recovery from a single RGB image. Predictions encompass the whole body, i.e., including hands and facial expressions, using the SMPL-X parametric model and 3D location in the camera coordinate system. Our model detects people by predicting coarse 2D heatmaps of person locations, using features produced by a standard Vision Transformer (ViT) backbone. It then predicts their whole-body pose, shape and 3D location using a new cross-attention module called the Human Prediction Head (HPH), with one query attending to the entire set of features for each detected person. As direct prediction of fine-grained hands and facial poses in a single shot, i.e., without relying on explicit crops around body parts, is hard to learn from existing data, we introduce CUFFS, the Close-Up Frames of Full-Body Subjects dataset, containing humans close to the camera with diverse hand poses. We show that incorporating it into the training data further enhances predictions, particularly for hands. Multi-HMR also optionally accounts for camera intrinsics, if available, by encoding camera ray directions for each image token. This simple design achieves strong performance on whole-body and body-only benchmarks simultaneously: a ViT-S backbone on $448{\times}448$ images already yields a fast and competitive model, while larger models and higher resolutions obtain state-of-the-art results.

Wonbong Jang, Philippe Weinzaepfel, Vincent Leroy et al.

We present Pow3r, a novel large 3D vision regression model that is highly versatile in the input modalities it accepts. Unlike previous feed-forward models that lack any mechanism to exploit known camera or scene priors at test time, Pow3r incorporates any combination of auxiliary information such as intrinsics, relative pose, dense or sparse depth, alongside input images, within a single network. Building upon the recent DUSt3R paradigm, a transformer-based architecture that leverages powerful pre-training, our lightweight and versatile conditioning acts as additional guidance for the network to predict more accurate estimates when auxiliary information is available. During training we feed the model with random subsets of modalities at each iteration, which enables the model to operate under different levels of known priors at test time. This in turn opens up new capabilities, such as performing inference in native image resolution, or point-cloud completion. Our experiments on 3D reconstruction, depth completion, multi-view depth prediction, multi-view stereo, and multi-view pose estimation tasks yield state-of-the-art results and confirm the effectiveness of Pow3r at exploiting all available information. The project webpage is https://europe.naverlabs.com/pow3r.

Mert Bulent Sariyildiz, Philippe Weinzaepfel, Thomas Lucas et al.

Recent multi-teacher distillation methods have unified the encoders of multiple foundation models into a single encoder, achieving competitive performance on core vision tasks like classification, segmentation, and depth estimation. This led us to ask: Could similar success be achieved when the pool of teachers also includes vision models specialized in diverse tasks across both 2D and 3D perception? In this paper, we define and investigate the problem of heterogeneous teacher distillation, or co-distillation, a challenging multi-teacher distillation scenario where teacher models vary significantly in both (a) their design objectives and (b) the data they were trained on. We explore data-sharing strategies and teacher-specific encoding, and introduce DUNE, a single encoder excelling in 2D vision, 3D understanding, and 3D human perception. Our model achieves performance comparable to that of its larger teachers, sometimes even outperforming them, on their respective tasks. Notably, DUNE surpasses MASt3R in Map-free Visual Relocalization with a much smaller encoder.

Nicolas Ugrinovic, Thomas Lucas, Fabien Baradel et al.

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.

Sara Rojas, Matthieu Armando, Bernard Ghamen et al.

Recovering the 3D geometry of a scene from a sparse set of uncalibrated images is a long-standing problem in computer vision. While recent learning-based approaches such as DUSt3R and MASt3R have demonstrated impressive results by directly predicting dense scene geometry, they are primarily trained on outdoor scenes with static environments and struggle to handle human-centric scenarios. In this work, we introduce HAMSt3R, an extension of MASt3R for joint human and scene 3D reconstruction from sparse, uncalibrated multi-view images. First, we exploit DUNE, a strong image encoder obtained by distilling, among others, the encoders from MASt3R and from a state-of-the-art Human Mesh Recovery (HMR) model, multi-HMR, for a better understanding of scene geometry and human bodies. Our method then incorporates additional network heads to segment people, estimate dense correspondences via DensePose, and predict depth in human-centric environments, enabling a more comprehensive 3D reconstruction. By leveraging the outputs of our different heads, HAMSt3R produces a dense point map enriched with human semantic information in 3D. Unlike existing methods that rely on complex optimization pipelines, our approach is fully feed-forward and efficient, making it suitable for real-world applications. We evaluate our model on EgoHumans and EgoExo4D, two challenging benchmarks con taining diverse human-centric scenarios. Additionally, we validate its generalization to traditional multi-view stereo and multi-view pose regression tasks. Our results demonstrate that our method can reconstruct humans effectively while preserving strong performance in general 3D reconstruction tasks, bridging the gap between human and scene understanding in 3D vision.

Gianluca Monaci, Philippe Weinzaepfel, Christian Wolf

Image goal navigation requires two different skills: firstly, core navigation skills, including the detection of free space and obstacles, and taking decisions based on an internal representation; and secondly, computing directional information by comparing visual observations to the goal image. Current state-of-the-art methods either rely on dedicated image-matching, or pre-training of computer vision modules on relative pose estimation. In this paper, we study whether this task can be efficiently solved with end-to-end training of full agents with RL, as has been claimed by recent work. A positive answer would have impact beyond Embodied AI and allow training of relative pose estimation from reward for navigation alone. In a large study we investigate the effect of architectural choices like late fusion, channel stacking, space-to-depth projections and cross-attention, and their role in the emergence of relative pose estimators from navigation training. We show that the success of recent methods is influenced up to a certain extent by simulator settings, leading to shortcuts in simulation. However, we also show that these capabilities can be transferred to more realistic setting, up to some extend. We also find evidence for correlations between navigation performance and probed (emerging) relative pose estimation performance, an important sub skill.

Mert Bulent Sariyildiz, Philippe Weinzaepfel, Guillaume Bono et al.

One key aspect of spatially aware robots is the ability to "find their bearings", ie. to correctly situate themselves in previously seen spaces. In this work, we focus on this particular scenario of continuous robotics operations, where information observed before an actual episode start is exploited to optimize efficiency. We introduce a new model, Kinaema, and agent, capable of integrating a stream of visual observations while moving in a potentially large scene, and upon request, processing a query image and predicting the relative position of the shown space with respect to its current position. Our model does not explicitly store an observation history, therefore does not have hard constraints on context length. It maintains an implicit latent memory, which is updated by a transformer in a recurrent way, compressing the history of sensor readings into a compact representation. We evaluate the impact of this model in a new downstream task we call "Mem-Nav". We show that our large-capacity recurrent model maintains a useful representation of the scene, navigates to goals observed before the actual episode start, and is computationally efficient, in particular compared to classical transformers with attention over an observation history.

Anilkumar Swamy, Vincent Leroy, Philippe Weinzaepfel et al.

Hand-object 3D reconstruction has become increasingly important for applications in human-robot interaction and immersive AR/VR experiences. A common approach for object-agnostic hand-object reconstruction from RGB sequences involves a two-stage pipeline: hand-object 3D tracking followed by multi-view 3D reconstruction. However, existing methods rely on keypoint detection techniques, such as Structure from Motion (SfM) and hand-keypoint optimization, which struggle with diverse object geometries, weak textures, and mutual hand-object occlusions, limiting scalability and generalization. As a key enabler to generic and seamless, non-intrusive applicability, we propose in this work a robust, keypoint detector-free approach to estimating hand-object 3D transformations from monocular motion video/images. We further integrate this with a multi-view reconstruction pipeline to accurately recover hand-object 3D shape. Our method, named HOSt3R, is unconstrained, does not rely on pre-scanned object templates or camera intrinsics, and reaches state-of-the-art performance for the tasks of object-agnostic hand-object 3D transformation and shape estimation on the SHOWMe benchmark. We also experiment on sequences from the HO3D dataset, demonstrating generalization to unseen object categories.

Yannis Kalantidis, Mert Bülent Sarıyıldız, Rafael S. Rezende et al.

State-of-the-art visual localization approaches generally rely on a first image retrieval step whose role is crucial. Yet, retrieval often struggles when facing varying conditions, due to e.g. weather or time of day, with dramatic consequences on the visual localization accuracy. In this paper, we improve this retrieval step and tailor it to the final localization task. Among the several changes we advocate for, we propose to synthesize variants of the training set images, obtained from generative text-to-image models, in order to automatically expand the training set towards a number of nameable variations that particularly hurt visual localization. After expanding the training set, we propose a training approach that leverages the specificities and the underlying geometry of this mix of real and synthetic images. We experimentally show that those changes translate into large improvements for the most challenging visual localization datasets. Project page: https://europe.naverlabs.com/ret4loc

Thomas Lucas, Philippe Weinzaepfel, Gregory Rogez

This paper tackles the problem of semi-supervised learning when the set of labeled samples is limited to a small number of images per class, typically less than 10, problem that we refer to as barely-supervised learning. We analyze in depth the behavior of a state-of-the-art semi-supervised method, FixMatch, which relies on a weakly-augmented version of an image to obtain supervision signal for a more strongly-augmented version. We show that it frequently fails in barely-supervised scenarios, due to a lack of training signal when no pseudo-label can be predicted with high confidence. We propose a method to leverage self-supervised methods that provides training signal in the absence of confident pseudo-labels. We then propose two methods to refine the pseudo-label selection process which lead to further improvements. The first one relies on a per-sample history of the model predictions, akin to a voting scheme. The second iteratively updates class-dependent confidence thresholds to better explore classes that are under-represented in the pseudo-labels. Our experiments show that our approach performs significantly better on STL-10 in the barely-supervised regime, e.g. with 4 or 8 labeled images per class.

Fabien Baradel, Thibault Groueix, Philippe Weinzaepfel et al.

Training state-of-the-art models for human body pose and shape recovery from images or videos requires datasets with corresponding annotations that are really hard and expensive to obtain. Our goal in this paper is to study whether poses from 3D Motion Capture (MoCap) data can be used to improve image-based and video-based human mesh recovery methods. We find that fine-tune image-based models with synthetic renderings from MoCap data can increase their performance, by providing them with a wider variety of poses, textures and backgrounds. In fact, we show that simply fine-tuning the batch normalization layers of the model is enough to achieve large gains. We further study the use of MoCap data for video, and introduce PoseBERT, a transformer module that directly regresses the pose parameters and is trained via masked modeling. It is simple, generic and can be plugged on top of any state-of-the-art image-based model in order to transform it in a video-based model leveraging temporal information. Our experimental results show that the proposed approaches reach state-of-the-art performance on various datasets including 3DPW, MPI-INF-3DHP, MuPoTS-3D, MCB and AIST. Test code and models will be available soon.

Donghwan Lee, Soohyun Ryu, Suyong Yeon et al.

Estimating the precise location of a camera using visual localization enables interesting applications such as augmented reality or robot navigation. This is particularly useful in indoor environments where other localization technologies, such as GNSS, fail. Indoor spaces impose interesting challenges on visual localization algorithms: occlusions due to people, textureless surfaces, large viewpoint changes, low light, repetitive textures, etc. Existing indoor datasets are either comparably small or do only cover a subset of the mentioned challenges. In this paper, we introduce 5 new indoor datasets for visual localization in challenging real-world environments. They were captured in a large shopping mall and a large metro station in Seoul, South Korea, using a dedicated mapping platform consisting of 10 cameras and 2 laser scanners. In order to obtain accurate ground truth camera poses, we developed a robust LiDAR SLAM which provides initial poses that are then refined using a novel structure-from-motion based optimization. We present a benchmark of modern visual localization algorithms on these challenging datasets showing superior performance of structure-based methods using robust image features. The datasets are available at: https://naverlabs.com/datasets

Vincent Leroy, Philippe Weinzaepfel, Romain Brégier et al.

Predicting 3D human pose from images has seen great recent improvements. Novel approaches that can even predict both pose and shape from a single input image have been introduced, often relying on a parametric model of the human body such as SMPL. While qualitative results for such methods are often shown for images captured in-the-wild, a proper benchmark in such conditions is still missing, as it is cumbersome to obtain ground-truth 3D poses elsewhere than in a motion capture room. This paper presents a pipeline to easily produce and validate such a dataset with accurate ground-truth, with which we benchmark recent 3D human pose estimation methods in-the-wild. We make use of the recently introduced Mannequin Challenge dataset which contains in-the-wild videos of people frozen in action like statues and leverage the fact that people are static and the camera moving to accurately fit the SMPL model on the sequences. A total of 24,428 frames with registered body models are then selected from 567 scenes at almost no cost, using only online RGB videos. We benchmark state-of-the-art SMPL-based human pose estimation methods on this dataset. Our results highlight that challenges remain, in particular for difficult poses or for scenes where the persons are partially truncated or occluded.

Yannis Kalantidis, Mert Bulent Sariyildiz, Noe Pion et al.

Contrastive learning has become a key component of self-supervised learning approaches for computer vision. By learning to embed two augmented versions of the same image close to each other and to push the embeddings of different images apart, one can train highly transferable visual representations. As revealed by recent studies, heavy data augmentation and large sets of negatives are both crucial in learning such representations. At the same time, data mixing strategies either at the image or the feature level improve both supervised and semi-supervised learning by synthesizing novel examples, forcing networks to learn more robust features. In this paper, we argue that an important aspect of contrastive learning, i.e., the effect of hard negatives, has so far been neglected. To get more meaningful negative samples, current top contrastive self-supervised learning approaches either substantially increase the batch sizes, or keep very large memory banks; increasing the memory size, however, leads to diminishing returns in terms of performance. We therefore start by delving deeper into a top-performing framework and show evidence that harder negatives are needed to facilitate better and faster learning. Based on these observations, and motivated by the success of data mixing, we propose hard negative mixing strategies at the feature level, that can be computed on-the-fly with a minimal computational overhead. We exhaustively ablate our approach on linear classification, object detection and instance segmentation and show that employing our hard negative mixing procedure improves the quality of visual representations learned by a state-of-the-art self-supervised learning method.

Philippe Weinzaepfel, Romain Brégier, Hadrien Combaluzier et al.

We introduce DOPE, the first method to detect and estimate whole-body 3D human poses, including bodies, hands and faces, in the wild. Achieving this level of details is key for a number of applications that require understanding the interactions of the people with each other or with the environment. The main challenge is the lack of in-the-wild data with labeled whole-body 3D poses. In previous work, training data has been annotated or generated for simpler tasks focusing on bodies, hands or faces separately. In this work, we propose to take advantage of these datasets to train independent experts for each part, namely a body, a hand and a face expert, and distill their knowledge into a single deep network designed for whole-body 2D-3D pose detection. In practice, given a training image with partial or no annotation, each part expert detects its subset of keypoints in 2D and 3D and the resulting estimations are combined to obtain whole-body pseudo ground-truth poses. A distillation loss encourages the whole-body predictions to mimic the experts' outputs. Our results show that this approach significantly outperforms the same whole-body model trained without distillation while staying close to the performance of the experts. Importantly, DOPE is computationally less demanding than the ensemble of experts and can achieve real-time performance. Test code and models are available at https://europe.naverlabs.com/research/computer-vision/dope.

Anil Armagan, Guillermo Garcia-Hernando, Seungryul Baek et al.

We study how well different types of approaches generalise in the task of 3D hand pose estimation under single hand scenarios and hand-object interaction. We show that the accuracy of state-of-the-art methods can drop, and that they fail mostly on poses absent from the training set. Unfortunately, since the space of hand poses is highly dimensional, it is inherently not feasible to cover the whole space densely, despite recent efforts in collecting large-scale training datasets. This sampling problem is even more severe when hands are interacting with objects and/or inputs are RGB rather than depth images, as RGB images also vary with lighting conditions and colors. To address these issues, we designed a public challenge (HANDS'19) to evaluate the abilities of current 3D hand pose estimators (HPEs) to interpolate and extrapolate the poses of a training set. More exactly, HANDS'19 is designed (a) to evaluate the influence of both depth and color modalities on 3D hand pose estimation, under the presence or absence of objects; (b) to assess the generalisation abilities w.r.t. four main axes: shapes, articulations, viewpoints, and objects; (c) to explore the use of a synthetic hand model to fill the gaps of current datasets. Through the challenge, the overall accuracy has dramatically improved over the baseline, especially on extrapolation tasks, from 27mm to 13mm mean joint error. Our analyses highlight the impacts of: Data pre-processing, ensemble approaches, the use of a parametric 3D hand model (MANO), and different HPE methods/backbones.

Philippe Weinzaepfel, Grégory Rogez

Recent methods for video action recognition have reached outstanding performances on existing benchmarks. However, they tend to leverage context such as scenes or objects instead of focusing on understanding the human action itself. For instance, a tennis field leads to the prediction playing tennis irrespectively of the actions performed in the video. In contrast, humans have a more complete understanding of actions and can recognize them without context. The best example of out-of-context actions are mimes, that people can typically recognize despite missing relevant objects and scenes. In this paper, we propose to benchmark action recognition methods in such absence of context and introduce a novel dataset, Mimetics, consisting of mimed actions for a subset of 50 classes from the Kinetics benchmark. Our experiments show that (a) state-of-the-art 3D convolutional neural networks obtain disappointing results on such videos, highlighting the lack of true understanding of the human actions and (b) models leveraging body language via human pose are less prone to context biases. In particular, we show that applying a shallow neural network with a single temporal convolution over body pose features transferred to the action recognition problem performs surprisingly well compared to 3D action recognition methods.

Jerome Revaud, Philippe Weinzaepfel, César De Souza et al.

Interest point detection and local feature description are fundamental steps in many computer vision applications. Classical methods for these tasks are based on a detect-then-describe paradigm where separate handcrafted methods are used to first identify repeatable keypoints and then represent them with a local descriptor. Neural networks trained with metric learning losses have recently caught up with these techniques, focusing on learning repeatable saliency maps for keypoint detection and learning descriptors at the detected keypoint locations. In this work, we argue that salient regions are not necessarily discriminative, and therefore can harm the performance of the description. Furthermore, we claim that descriptors should be learned only in regions for which matching can be performed with high confidence. We thus propose to jointly learn keypoint detection and description together with a predictor of the local descriptor discriminativeness. This allows us to avoid ambiguous areas and leads to reliable keypoint detections and descriptions. Our detection-and-description approach, trained with self-supervision, can simultaneously output sparse, repeatable and reliable keypoints that outperforms state-of-the-art detectors and descriptors on the HPatches dataset. It also establishes a record on the recently released Aachen Day-Night localization dataset.

Gregory Rogez, Philippe Weinzaepfel, Cordelia Schmid

We propose an end-to-end architecture for joint 2D and 3D human pose estimation in natural images. Key to our approach is the generation and scoring of a number of pose proposals per image, which allows us to predict 2D and 3D poses of multiple people simultaneously. Hence, our approach does not require an approximate localization of the humans for initialization. Our Localization-Classification-Regression architecture, named LCR-Net, contains 3 main components: 1) the pose proposal generator that suggests candidate poses at different locations in the image; 2) a classifier that scores the different pose proposals; and 3) a regressor that refines pose proposals both in 2D and 3D. All three stages share the convolutional feature layers and are trained jointly. The final pose estimation is obtained by integrating over neighboring pose hypotheses, which is shown to improve over a standard non maximum suppression algorithm. Our method recovers full-body 2D and 3D poses, hallucinating plausible body parts when the persons are partially occluded or truncated by the image boundary. Our approach significantly outperforms the state of the art in 3D pose estimation on Human3.6M, a controlled environment. Moreover, it shows promising results on real images for both single and multi-person subsets of the MPII 2D pose benchmark and demonstrates satisfying 3D pose results even for multi-person images.

Vicky Kalogeiton, Philippe Weinzaepfel, Vittorio Ferrari et al.

Current state-of-the-art approaches for spatio-temporal action localization rely on detections at the frame level that are then linked or tracked across time. In this paper, we leverage the temporal continuity of videos instead of operating at the frame level. We propose the ACtion Tubelet detector (ACT-detector) that takes as input a sequence of frames and outputs tubelets, i.e., sequences of bounding boxes with associated scores. The same way state-of-the-art object detectors rely on anchor boxes, our ACT-detector is based on anchor cuboids. We build upon the SSD framework. Convolutional features are extracted for each frame, while scores and regressions are based on the temporal stacking of these features, thus exploiting information from a sequence. Our experimental results show that leveraging sequences of frames significantly improves detection performance over using individual frames. The gain of our tubelet detector can be explained by both more accurate scores and more precise localization. Our ACT-detector outperforms the state-of-the-art methods for frame-mAP and video-mAP on the J-HMDB and UCF-101 datasets, in particular at high overlap thresholds.

Philippe Weinzaepfel, Xavier Martin, Cordelia Schmid

We introduce an approach for spatio-temporal human action localization using sparse spatial supervision. Our method leverages the large amount of annotated humans available today and extracts human tubes by combining a state-of-the-art human detector with a tracking-by-detection approach. Given these high-quality human tubes and temporal supervision, we select positive and negative tubes with very sparse spatial supervision, i.e., only one spatially annotated frame per instance. The selected tubes allow us to effectively learn a spatio-temporal action detector based on dense trajectories or CNNs. We conduct experiments on existing action localization benchmarks: UCF-Sports, J-HMDB and UCF-101. Our results show that our approach, despite using sparse spatial supervision, performs on par with methods using full supervision, i.e., one bounding box annotation per frame. To further validate our method, we introduce DALY (Daily Action Localization in YouTube), a dataset for realistic action localization in space and time. It contains high quality temporal and spatial annotations for 3.6k instances of 10 actions in 31 hours of videos (3.3M frames). It is an order of magnitude larger than existing datasets, with more diversity in appearance and long untrimmed videos.

Jerome Revaud, Philippe Weinzaepfel, Zaid Harchaoui et al.

We introduce a novel matching algorithm, called DeepMatching, to compute dense correspondences between images. DeepMatching relies on a hierarchical, multi-layer, correlational architecture designed for matching images and was inspired by deep convolutional approaches. The proposed matching algorithm can handle non-rigid deformations and repetitive textures and efficiently determines dense correspondences in the presence of significant changes between images. We evaluate the performance of DeepMatching, in comparison with state-of-the-art matching algorithms, on the Mikolajczyk (Mikolajczyk et al 2005), the MPI-Sintel (Butler et al 2012) and the Kitti (Geiger et al 2013) datasets. DeepMatching outperforms the state-of-the-art algorithms and shows excellent results in particular for repetitive textures.We also propose a method for estimating optical flow, called DeepFlow, by integrating DeepMatching in the large displacement optical flow (LDOF) approach of Brox and Malik (2011). Compared to existing matching algorithms, additional robustness to large displacements and complex motion is obtained thanks to our matching approach. DeepFlow obtains competitive performance on public benchmarks for optical flow estimation.

Philippe Weinzaepfel, Zaid Harchaoui, Cordelia Schmid

We propose an effective approach for spatio-temporal action localization in realistic videos. The approach first detects proposals at the frame-level and scores them with a combination of static and motion CNN features. It then tracks high-scoring proposals throughout the video using a tracking-by-detection approach. Our tracker relies simultaneously on instance-level and class-level detectors. The tracks are scored using a spatio-temporal motion histogram, a descriptor at the track level, in combination with the CNN features. Finally, we perform temporal localization of the action using a sliding-window approach at the track level. We present experimental results for spatio-temporal localization on the UCF-Sports, J-HMDB and UCF-101 action localization datasets, where our approach outperforms the state of the art with a margin of 15%, 7% and 12% respectively in mAP.

Jerome Revaud, Philippe Weinzaepfel, Zaid Harchaoui et al.

We propose a novel approach for optical flow estimation , targeted at large displacements with significant oc-clusions. It consists of two steps: i) dense matching by edge-preserving interpolation from a sparse set of matches; ii) variational energy minimization initialized with the dense matches. The sparse-to-dense interpolation relies on an appropriate choice of the distance, namely an edge-aware geodesic distance. This distance is tailored to handle occlusions and motion boundaries -- two common and difficult issues for optical flow computation. We also propose an approximation scheme for the geodesic distance to allow fast computation without loss of performance. Subsequent to the dense interpolation step, standard one-level variational energy minimization is carried out on the dense matches to obtain the final flow estimation. The proposed approach, called Edge-Preserving Interpolation of Correspondences (EpicFlow) is fast and robust to large displacements. It significantly outperforms the state of the art on MPI-Sintel and performs on par on Kitti and Middlebury.