Hyeongwoo Kim, Pablo Garrido, Ayush Tewari, Weipeng Xu, Justus Thies, Matthias Nießner, Patrick Pérez, Christian Richardt, Michael Zollhöfer, Christian Theobalt
We present a novel approach that enables photo-realistic re-animation of portrait videos using only an input video. In contrast to existing approaches that are restricted to manipulations of facial expressions only, we are the first to transfer the full 3D head position, head rotation, face expression, eye gaze, and eye blinking from a source actor to a portrait video of a target actor. The core of our approach is a generative neural network with a novel space-time architecture. The network takes as input synthetic renderings of a parametric face model, based on which it predicts photo-realistic video frames for a given target actor. The realism in this rendering-to-video transfer is achieved by careful adversarial training, and as a result, we can create modified target videos that mimic the behavior of the synthetically-created input. In order to enable source-to-target video re-animation, we render a synthetic target video with the reconstructed head animation parameters from a source video, and feed it into the trained network -- thus taking full control of the target. With the ability to freely recombine source and target parameters, we are able to demonstrate a large variety of video rewrite applications without explicitly modeling hair, body or background. For instance, we can reenact the full head using interactive user-controlled editing, and realize high-fidelity visual dubbing. To demonstrate the high quality of our output, we conduct an extensive series of experiments and evaluations, where for instance a user study shows that our video edits are hard to detect.
Maximilian Jaritz, Tuan-Hung Vu, Raoul de Charette, Émilie Wirbel, Patrick Pérez
Unsupervised Domain Adaptation (UDA) is crucial to tackle the lack of annotations in a new domain. There are many multi-modal datasets, but most UDA approaches are uni-modal. In this work, we explore how to learn from multi-modality and propose cross-modal UDA (xMUDA) where we assume the presence of 2D images and 3D point clouds for 3D semantic segmentation. This is challenging as the two input spaces are heterogeneous and can be impacted differently by domain shift. In xMUDA, modalities learn from each other through mutual mimicking, disentangled from the segmentation objective, to prevent the stronger modality from adopting false predictions from the weaker one. We evaluate on new UDA scenarios including day-to-night, country-to-country and dataset-to-dataset, leveraging recent autonomous driving datasets. xMUDA brings large improvements over uni-modal UDA on all tested scenarios, and is complementary to state-of-the-art UDA techniques. Code is available at https://github.com/valeoai/xmuda.
Spyros Gidaris, Andrei Bursuc, Nikos Komodakis, Patrick Pérez, Matthieu Cord
Few-shot learning and self-supervised learning address different facets of the same problem: how to train a model with little or no labeled data. Few-shot learning aims for optimization methods and models that can learn efficiently to recognize patterns in the low data regime. Self-supervised learning focuses instead on unlabeled data and looks into it for the supervisory signal to feed high capacity deep neural networks. In this work we exploit the complementarity of these two domains and propose an approach for improving few-shot learning through self-supervision. We use self-supervision as an auxiliary task in a few-shot learning pipeline, enabling feature extractors to learn richer and more transferable visual representations while still using few annotated samples. Through self-supervision, our approach can be naturally extended towards using diverse unlabeled data from other datasets in the few-shot setting. We report consistent improvements across an array of architectures, datasets and self-supervision techniques.
Simon Roburin, Yann de Mont-Marin, Andrei Bursuc, Renaud Marlet, Patrick Pérez, Mathieu Aubry
Normalization Layers (NLs) are widely used in modern deep-learning architectures. Despite their apparent simplicity, their effect on optimization is not yet fully understood. This paper introduces a spherical framework to study the optimization of neural networks with NLs from a geometric perspective. Concretely, the radial invariance of groups of parameters, such as filters for convolutional neural networks, allows to translate the optimization steps on the $L_2$ unit hypersphere. This formulation and the associated geometric interpretation shed new light on the training dynamics. Firstly, the first effective learning rate expression of Adam is derived. Then the demonstration that, in the presence of NLs, performing Stochastic Gradient Descent (SGD) alone is actually equivalent to a variant of Adam constrained to the unit hypersphere, stems from the framework. Finally, this analysis outlines phenomena that previous variants of Adam act on and their importance in the optimization process are experimentally validated.
Alasdair Newson, Andrés Almansa, Matthieu Fradet, Yann Gousseau, Patrick Pérez
We propose an automatic video inpainting algorithm which relies on the optimisation of a global, patch-based functional. Our algorithm is able to deal with a variety of challenging situations which naturally arise in video inpainting, such as the correct reconstruction of dynamic textures, multiple moving objects and moving background. Furthermore, we achieve this in an order of magnitude less execution time with respect to the state-of-the-art. We are also able to achieve good quality results on high definition videos. Finally, we provide specific algorithmic details to make implementation of our algorithm as easy as possible. The resulting algorithm requires no segmentation or manual input other than the definition of the inpainting mask, and can deal with a wider variety of situations than is handled by previous work. 1. Introduction. Advanced image and video editing techniques are increasingly common in the image processing and computer vision world, and are also starting to be used in media entertainment. One common and difficult task closely linked to the world of video editing is image and video " inpainting ". Generally speaking, this is the task of replacing the content of an image or video with some other content which is visually pleasing. This subject has been extensively studied in the case of images, to such an extent that commercial image inpainting products destined for the general public are available, such as Photoshop's " Content Aware fill " [1]. However, while some impressive results have been obtained in the case of videos, the subject has been studied far less extensively than image inpainting. This relative lack of research can largely be attributed to high time complexity due to the added temporal dimension. Indeed, it has only very recently become possible to produce good quality inpainting results on high definition videos, and this only in a semi-automatic manner. Nevertheless, high-quality video inpainting has many important and useful applications such as film restoration, professional post-production in cinema and video editing for personal use. For this reason, we believe that an automatic, generic video inpainting algorithm would be extremely useful for both academic and professional communities.
Eric Grinstein, Ngoc Duong, Alexey Ozerov, Patrick Pérez
'Style transfer' among images has recently emerged as a very active research topic, fuelled by the power of convolution neural networks (CNNs), and has become fast a very popular technology in social media. This paper investigates the analogous problem in the audio domain: How to transfer the style of a reference audio signal to a target audio content? We propose a flexible framework for the task, which uses a sound texture model to extract statistics characterizing the reference audio style, followed by an optimization-based audio texture synthesis to modify the target content. In contrast to mainstream optimization-based visual transfer method, the proposed process is initialized by the target content instead of random noise and the optimized loss is only about texture, not structure. These differences proved key for audio style transfer in our experiments. In order to extract features of interest, we investigate different architectures, whether pre-trained on other tasks, as done in image style transfer, or engineered based on the human auditory system. Experimental results on different types of audio signal confirm the potential of the proposed approach.
Hippolyte Pilchen, Edouard Grave, Patrick Pérez
Recent techniques such as retrieval-augmented generation or chain-of-thought reasoning have led to longer contexts and increased inference costs. Context compression techniques can reduce these costs, but the most effective approaches require fine-tuning the target model or even modifying its architecture. This can degrade its general abilities when not used for this specific purpose. Here we explore an alternative approach: an encoder that compresses the context into continuous representations which replace token embeddings in decoder LLMs. First, we perform a systematic study of training strategies and architecture choices for the encoder. Our findings led to the design of an Adaptable text Representations Compressor, named ARC-Encoder, which outputs $x$-times fewer continuous representations (typically $x\!\in\!\{4,8\}$) than text tokens. We evaluate ARC-Encoder across a variety of LLM usage scenarios, ranging from in-context learning to context window extension, on both instruct and base decoders. Results show that ARC-Encoder achieves state-of-the-art performance on several benchmarks while improving computational efficiency at inference. Finally, we demonstrate that our models can be adapted to multiple decoders simultaneously, allowing a single encoder to generalize across different decoder LLMs. This makes ARC-Encoder a flexible and efficient solution for portable encoders that work seamlessly with multiple LLMs. We release a training code at https://github.com/kyutai-labs/ARC-Encoder , fine-tuning dataset and pretrained models are available at https://huggingface.co/collections/kyutai/arc-encoders-68ee18787301407d60a57047 .
Victor Besnier, Himalaya Jain, Andrei Bursuc, Matthieu Cord, Patrick Pérez
Current generative networks are increasingly proficient in generating high-resolution realistic images. These generative networks, especially the conditional ones, can potentially become a great tool for providing new image datasets. This naturally brings the question: Can we train a classifier only on the generated data? This potential availability of nearly unlimited amounts of training data challenges standard practices for training machine learning models, which have been crafted across the years for limited and fixed size datasets. In this work we investigate this question and its related challenges. We identify ways to improve significantly the performance over naive training on randomly generated images with regular heuristics. We propose three standalone techniques that can be applied at different stages of the pipeline, i.e., data generation, training on generated data, and deploying on real data. We evaluate our proposed approaches on a subset of the ImageNet dataset and show encouraging results compared to classifiers trained on real images.
Charles Corbière, Nicolas Thome, Antoine Saporta, Tuan-Hung Vu, Matthieu Cord, Patrick Pérez
Reliably quantifying the confidence of deep neural classifiers is a challenging yet fundamental requirement for deploying such models in safety-critical applications. In this paper, we introduce a novel target criterion for model confidence, namely the true class probability (TCP). We show that TCP offers better properties for confidence estimation than standard maximum class probability (MCP). Since the true class is by essence unknown at test time, we propose to learn TCP criterion from data with an auxiliary model, introducing a specific learning scheme adapted to this context. We evaluate our approach on the task of failure prediction and of self-training with pseudo-labels for domain adaptation, which both necessitate effective confidence estimates. Extensive experiments are conducted for validating the relevance of the proposed approach in each task. We study various network architectures and experiment with small and large datasets for image classification and semantic segmentation. In every tested benchmark, our approach outperforms strong baselines.
Julien Rebut, Andrei Bursuc, Patrick Pérez
Deep Neural Networks (DNNs) are a critical component for self-driving vehicles. They achieve impressive performance by reaping information from high amounts of labeled data. Yet, the full complexity of the real world cannot be encapsulated in the training data, no matter how big the dataset, and DNNs can hardly generalize to unseen conditions. Robustness to various image corruptions, caused by changing weather conditions or sensor degradation and aging, is crucial for safety when such vehicles are deployed in the real world. We address this problem through a novel type of layer, dubbed StyleLess, which enables DNNs to learn robust and informative features that can cope with varying external conditions. We propose multiple variations of this layer that can be integrated in most of the architectures and trained jointly with the main task. We validate our contribution on typical autonomous-driving tasks (detection, semantic segmentation), showing that in most cases, this approach improves predictive performance on unseen conditions (fog, rain), while preserving performance on seen conditions and objects.
Oriane Siméoni, Gilles Puy, Huy V. Vo, Simon Roburin, Spyros Gidaris, Andrei Bursuc, Patrick Pérez, Renaud Marlet, Jean Ponce
Localizing objects in image collections without supervision can help to avoid expensive annotation campaigns. We propose a simple approach to this problem, that leverages the activation features of a vision transformer pre-trained in a self-supervised manner. Our method, LOST, does not require any external object proposal nor any exploration of the image collection; it operates on a single image. Yet, we outperform state-of-the-art object discovery methods by up to 8 CorLoc points on PASCAL VOC 2012. We also show that training a class-agnostic detector on the discovered objects boosts results by another 7 points. Moreover, we show promising results on the unsupervised object discovery task. The code to reproduce our results can be found at https://github.com/valeoai/LOST.
Huy V. Vo, Ngoc Q. K. Duong, Patrick Perez
Scene-agnostic visual inpainting remains very challenging despite progress in patch-based methods. Recently, Pathak et al. 2016 have introduced convolutional "context encoders" (CEs) for unsupervised feature learning through image completion tasks. With the additional help of adversarial training, CEs turned out to be a promising tool to complete complex structures in real inpainting problems. In the present paper we propose to push further this key ability by relying on perceptual reconstruction losses at training time. We show on a wide variety of visual scenes the merit of the approach for structural inpainting, and confirm it through a user study. Combined with the optimization-based refinement of Yang et al. 2016 with neural patches, our context encoder opens up new opportunities for prior-free visual inpainting.
Ayush Tewari, Michael Zollhöfer, Pablo Garrido, Florian Bernard, Hyeongwoo Kim, Patrick Pérez, Christian Theobalt
The reconstruction of dense 3D models of face geometry and appearance from a single image is highly challenging and ill-posed. To constrain the problem, many approaches rely on strong priors, such as parametric face models learned from limited 3D scan data. However, prior models restrict generalization of the true diversity in facial geometry, skin reflectance and illumination. To alleviate this problem, we present the first approach that jointly learns 1) a regressor for face shape, expression, reflectance and illumination on the basis of 2) a concurrently learned parametric face model. Our multi-level face model combines the advantage of 3D Morphable Models for regularization with the out-of-space generalization of a learned corrective space. We train end-to-end on in-the-wild images without dense annotations by fusing a convolutional encoder with a differentiable expert-designed renderer and a self-supervised training loss, both defined at multiple detail levels. Our approach compares favorably to the state-of-the-art in terms of reconstruction quality, better generalizes to real world faces, and runs at over 250 Hz.
Gilles Puy, Spyros Gidaris, Alexandre Boulch, Oriane Siméoni, Corentin Sautier, Patrick Pérez, Andrei Bursuc, Renaud Marlet
Self-supervised image backbones can be used to address complex 2D tasks (e.g., semantic segmentation, object discovery) very efficiently and with little or no downstream supervision. Ideally, 3D backbones for lidar should be able to inherit these properties after distillation of these powerful 2D features. The most recent methods for image-to-lidar distillation on autonomous driving data show promising results, obtained thanks to distillation methods that keep improving. Yet, we still notice a large performance gap when measuring the quality of distilled and fully supervised features by linear probing. In this work, instead of focusing only on the distillation method, we study the effect of three pillars for distillation: the 3D backbone, the pretrained 2D backbones, and the pretraining dataset. In particular, thanks to our scalable distillation method named ScaLR, we show that scaling the 2D and 3D backbones and pretraining on diverse datasets leads to a substantial improvement of the feature quality. This allows us to significantly reduce the gap between the quality of distilled and fully-supervised 3D features, and to improve the robustness of the pretrained backbones to domain gaps and perturbations.
Gilles Puy, Patrick Pérez
We propose a new flexible deep convolutional neural network (convnet) to perform fast visual style transfer. In contrast to existing convnets that address the same task, our architecture derives directly from the structure of the gradient descent originally used to solve the style transfer problem [Gatys et al., 2016]. Like existing convnets, ours approximately solves the original problem much faster than the gradient descent. However, our network is uniquely flexible by design: it can be manipulated at runtime to enforce new constraints on the final solution. In particular, we show how to modify it to obtain a photorealistic result with no retraining. We study the modifications made by [Luan et al., 2017] to the original cost function of [Gatys et al., 2016] to achieve photorealistic style transfer. These modifications affect directly the gradient descent and can be reported on-the-fly in our network. These modifications are possible as the proposed architecture stems from unrolling the gradient descent.
Antonín Vobecký, David Hurych, Michal Uřičář, Patrick Pérez, Josef Šivic
Existing datasets for training pedestrian detectors in images suffer from limited appearance and pose variation. The most challenging scenarios are rarely included because they are too difficult to capture due to safety reasons, or they are very unlikely to happen. The strict safety requirements in assisted and autonomous driving applications call for an extra high detection accuracy also in these rare situations. Having the ability to generate people images in arbitrary poses, with arbitrary appearances and embedded in different background scenes with varying illumination and weather conditions, is a crucial component for the development and testing of such applications. The contributions of this paper are three-fold. First, we describe an augmentation method for controlled synthesis of urban scenes containing people, thus producing rare or never-seen situations. This is achieved with a data generator (called DummyNet) with disentangled control of the pose, the appearance, and the target background scene. Second, the proposed generator relies on novel network architecture and associated loss that takes into account the segmentation of the foreground person and its composition into the background scene. Finally, we demonstrate that the data generated by our DummyNet improve performance of several existing person detectors across various datasets as well as in challenging situations, such as night-time conditions, where only a limited amount of training data is available. In the setup with only day-time data available, we improve the night-time detector by $17\%$ log-average miss rate over the detector trained with the day-time data only.
Huy V. Vo, Elena Sizikova, Cordelia Schmid, Patrick Pérez, Jean Ponce
Existing approaches to unsupervised object discovery (UOD) do not scale up to large datasets without approximations that compromise their performance. We propose a novel formulation of UOD as a ranking problem, amenable to the arsenal of distributed methods available for eigenvalue problems and link analysis. Through the use of self-supervised features, we also demonstrate the first effective fully unsupervised pipeline for UOD. Extensive experiments on COCO and OpenImages show that, in the single-object discovery setting where a single prominent object is sought in each image, the proposed LOD (Large-scale Object Discovery) approach is on par with, or better than the state of the art for medium-scale datasets (up to 120K images), and over 37% better than the only other algorithms capable of scaling up to 1.7M images. In the multi-object discovery setting where multiple objects are sought in each image, the proposed LOD is over 14% better in average precision (AP) than all other methods for datasets ranging from 20K to 1.7M images. Using self-supervised features, we also show that the proposed method obtains state-of-the-art UOD performance on OpenImages. Our code is publicly available at https://github.com/huyvvo/LOD.
Guillaume Le Moing, Tuan-Hung Vu, Himalaya Jain, Patrick Pérez, Matthieu Cord
Despite the recent progress of generative adversarial networks (GANs) at synthesizing photo-realistic images, producing complex urban scenes remains a challenging problem. Previous works break down scene generation into two consecutive phases: unconditional semantic layout synthesis and image synthesis conditioned on layouts. In this work, we propose to condition layout generation as well for higher semantic control: given a vector of class proportions, we generate layouts with matching composition. To this end, we introduce a conditional framework with novel architecture designs and learning objectives, which effectively accommodates class proportions to guide the scene generation process. The proposed architecture also allows partial layout editing with interesting applications. Thanks to the semantic control, we can produce layouts close to the real distribution, helping enhance the whole scene generation process. On different metrics and urban scene benchmarks, our models outperform existing baselines. Moreover, we demonstrate the merit of our approach for data augmentation: semantic segmenters trained on real layout-image pairs along with additional ones generated by our approach outperform models only trained on real pairs.
Himalaya Jain, Joaquin Zepeda, Patrick Pérez, Rémi Gribonval
For large-scale visual search, highly compressed yet meaningful representations of images are essential. Structured vector quantizers based on product quantization and its variants are usually employed to achieve such compression while minimizing the loss of accuracy. Yet, unlike binary hashing schemes, these unsupervised methods have not yet benefited from the supervision, end-to-end learning and novel architectures ushered in by the deep learning revolution. We hence propose herein a novel method to make deep convolutional neural networks produce supervised, compact, structured binary codes for visual search. Our method makes use of a novel block-softmax non-linearity and of batch-based entropy losses that together induce structure in the learned encodings. We show that our method outperforms state-of-the-art compact representations based on deep hashing or structured quantization in single and cross-domain category retrieval, instance retrieval and classification. We make our code and models publicly available online.
Gilles Puy, Srdan Kitic, Patrick Pérez
This paper deals with the unification of local and non-local signal processing on graphs within a single convolutional neural network (CNN) framework. Building upon recent works on graph CNNs, we propose to use convolutional layers that take as inputs two variables, a signal and a graph, allowing the network to adapt to changes in the graph structure. In this article, we explain how this framework allows us to design a novel method to perform style transfer.