Search

In this work, we tackle the problem of learning universal robotic dexterous grasping from a point cloud observation under a table-top setting. The goal is to grasp and lift up objects in high-quality and diverse ways and generalize across hundreds of categories and even the unseen. Inspired by successful pipelines used in parallel gripper grasping, we split the task into two stages: 1) grasp proposal (pose) generation and 2) goal-conditioned grasp execution. For the first stage, we propose a novel probabilistic model of grasp pose conditioned on the point cloud observation that factorizes rotation from translation and articulation. Trained on our synthesized large-scale dexterous grasp dataset, this model enables us to sample diverse and high-quality dexterous grasp poses for the object point cloud.For the second stage, we propose to replace the motion planning used in parallel gripper grasping with a goal-conditioned grasp policy, due to the complexity involved in dexterous grasping execution. Note that it is very challenging to learn this highly generalizable grasp policy that only takes realistic inputs without oracle states. We thus propose several important innovations, including state canonicalization, object curriculum, and teacher-student distillation. Integrating the two stages, our final pipeline becomes the first to achieve universal generalization for dexterous grasping, demonstrating an average success rate of more than 60\% on thousands of object instances, which significantly outperforms all baselines, meanwhile showing only a minimal generalization gap., Comment: Accepted to CVPR 2023

The performance of the keyword spotting (KWS) system based on audio modality, commonly measured in false alarms and false rejects, degrades significantly under the far field and noisy conditions. Therefore, audio-visual keyword spotting, which leverages complementary relationships over multiple modalities, has recently gained much attention. However, current studies mainly focus on combining the exclusively learned representations of different modalities, instead of exploring the modal relationships during each respective modeling. In this paper, we propose a novel visual modality enhanced end-to-end KWS framework (VE-KWS), which fuses audio and visual modalities from two aspects. The first one is utilizing the speaker location information obtained from the lip region in videos to assist the training of multi-channel audio beamformer. By involving the beamformer as an audio enhancement module, the acoustic distortions, caused by the far field or noisy environments, could be significantly suppressed. The other one is conducting cross-attention between different modalities to capture the inter-modal relationships and help the representation learning of each modality. Experiments on the MSIP challenge corpus show that our proposed model achieves 2.79% false rejection rate and 2.95% false alarm rate on the Eval set, resulting in a new SOTA performance compared with the top-ranking systems in the ICASSP2022 MISP challenge., Comment: 5 pages. Accepted at ICASSP2023

Category-level articulated object pose estimation aims to estimate a hierarchy of articulation-aware object poses of an unseen articulated object from a known category. To reduce the heavy annotations needed for supervised learning methods, we present a novel self-supervised strategy that solves this problem without any human labels. Our key idea is to factorize canonical shapes and articulated object poses from input articulated shapes through part-level equivariant shape analysis. Specifically, we first introduce the concept of part-level SE(3) equivariance and devise a network to learn features of such property. Then, through a carefully designed fine-grained pose-shape disentanglement strategy, we expect that canonical spaces to support pose estimation could be induced automatically. Thus, we could further predict articulated object poses as per-part rigid transformations describing how parts transform from their canonical part spaces to the camera space. Extensive experiments demonstrate the effectiveness of our method on both complete and partial point clouds from synthetic and real articulated object datasets., Comment: ICLR 2023

We show that a local non-Hermitian perturbation in a Hermitian lattice system generically induces scale-free localization for the continuous-spectrum eigenstates. When the perturbation lies at a finite distance to the boundary, the scale-free eigenstates are promoted to exponentially localized modes, whose number is proportional to the distance. Furthermore, when the local non-Hermitian perturbation respects parity-time (PT) symmetry, the PT symmetry breaking is always accompanied by the emergence of scale-free or exponential localization. Intriguingly, we find a concise band-structure condition, which tells not only when the continuous-spectrum PT breaking of scale-free modes can occur, but also the precise PT-breaking energy window. Our results uncover a series of unexpected generic phenomena induced by a local non-Hermitian perturbation, which has interesting interplay with PT symmetry., Comment: main text: 6 pages, 4 figures; supplemental material: 14 pages, 7 figures

The early detection of glaucoma is essential in preventing visual impairment. Artificial intelligence (AI) can be used to analyze color fundus photographs (CFPs) in a cost-effective manner, making glaucoma screening more accessible. While AI models for glaucoma screening from CFPs have shown promising results in laboratory settings, their performance decreases significantly in real-world scenarios due to the presence of out-of-distribution and low-quality images. To address this issue, we propose the Artificial Intelligence for Robust Glaucoma Screening (AIROGS) challenge. This challenge includes a large dataset of around 113,000 images from about 60,000 patients and 500 different screening centers, and encourages the development of algorithms that are robust to ungradable and unexpected input data. We evaluated solutions from 14 teams in this paper, and found that the best teams performed similarly to a set of 20 expert ophthalmologists and optometrists. The highest-scoring team achieved an area under the receiver operating characteristic curve of 0.99 (95% CI: 0.98-0.99) for detecting ungradable images on-the-fly. Additionally, many of the algorithms showed robust performance when tested on three other publicly available datasets. These results demonstrate the feasibility of robust AI-enabled glaucoma screening., Comment: 19 pages, 8 figures, 3 tables

We study the open XXZ spin chain with a PT-symmetric non-Hermitian boundary field. We find an interaction-induced scale-free non-Hermitian skin effect by using the coordinate Bethe ansatz. The steady state and the ground state in the PT broken phase are constructed, and the formulas of their eigen-energies in the thermodynamic limit are obtained. The differences between the many-body scale-free states and the boundary string states are explored, and the transition between the two at isotropic point is investigated. We also discuss an experimental scheme to verify our results., Comment: 25 pages, 8 figures

We study the entanglement and work statistics in a driven two-qubit system. The regulation of periodic driving has much more versatility and universality in contrast to reservoir engineering in static systems. We found the quasi-steady state entanglement can be amplified effectively by the external drive in certain parameter regimes. The drive extends the range of temperatures or temperature differences at which entanglement can emerge. From the view of the effective Hamiltonian, the addition of the driving alters the inter-qubit coupling and system-bath coupling, which are crucial in determining the quasi-steady state. The work statistics are also investigated. The driven system, as a continuous quantum thermal machine, output work continuously and steadily at the quasi-steady state. There is a distinct operation of modes and corresponding performance by changing driving. It can also be understood that the drive changes the effective Hamiltonian, and further the modes of energy exchanges between the system and the baths as well as the work reservoir.

With the advent of gravitational-wave astronomy and the discovery of more compact binary coalescences, data quality improvement techniques are desired to handle the complex and overwhelming noise in gravitational wave (GW) observational data. Though recent machine learning-based studies have shown promising results for data denoising, they are unable to precisely recover both the GW signal amplitude and phase. To address such an issue, we develop a deep neural network centered workflow, WaveFormer, for significant noise suppression and signal recovery on observational data from the Laser Interferometer Gravitational-Wave Observatory (LIGO). The WaveFormer has a science-driven architecture design with hierarchical feature extraction across a broad frequency spectrum. As a result, the overall noise and glitch are decreased by more than one order of magnitude and the signal recovery error is roughly 1% and 7% for the phase and amplitude, respectively. Moreover, on 75 reported binary black hole (BBH) events of LIGO we obtain a significant improvement of inverse false alarm rate. Our work highlights the potential of large neural networks in gravitational wave data analysis and, while primarily demonstrated on LIGO data, its adaptable design indicates promise for broader application within the International Gravitational-Wave Observatories Network (IGWN) in future observational runs.

Fine-grained capturing of 3D HOI boosts human activity understanding and facilitates downstream visual tasks, including action recognition, holistic scene reconstruction, and human motion synthesis. Despite its significance, existing works mostly assume that humans interact with rigid objects using only a few body parts, limiting their scope. In this paper, we address the challenging problem of f-AHOI, wherein the whole human bodies interact with articulated objects, whose parts are connected by movable joints. We present CHAIRS, a large-scale motion-captured f-AHOI dataset, consisting of 16.2 hours of versatile interactions between 46 participants and 81 articulated and rigid sittable objects. CHAIRS provides 3D meshes of both humans and articulated objects during the entire interactive process, as well as realistic and physically plausible full-body interactions. We show the value of CHAIRS with object pose estimation. By learning the geometrical relationships in HOI, we devise the very first model that leverage human pose estimation to tackle the estimation of articulated object poses and shapes during whole-body interactions. Given an image and an estimated human pose, our model first reconstructs the pose and shape of the object, then optimizes the reconstruction according to a learned interaction prior. Under both evaluation settings (e.g., with or without the knowledge of objects' geometries/structures), our model significantly outperforms baselines. We hope CHAIRS will promote the community towards finer-grained interaction understanding. We will make the data/code publicly available.

Object goal navigation (ObjectNav) in unseen environments is a fundamental task for Embodied AI. Agents in existing works learn ObjectNav policies based on 2D maps, scene graphs, or image sequences. Considering this task happens in 3D space, a 3D-aware agent can advance its ObjectNav capability via learning from fine-grained spatial information. However, leveraging 3D scene representation can be prohibitively unpractical for policy learning in this floor-level task, due to low sample efficiency and expensive computational cost. In this work, we propose a framework for the challenging 3D-aware ObjectNav based on two straightforward sub-policies. The two sub-polices, namely corner-guided exploration policy and category-aware identification policy, simultaneously perform by utilizing online fused 3D points as observation. Through extensive experiments, we show that this framework can dramatically improve the performance in ObjectNav through learning from 3D scene representation. Our framework achieves the best performance among all modular-based methods on the Matterport3D and Gibson datasets, while requiring (up to 30x) less computational cost for training., Comment: To appear in CVPR 2023

Due to the lack of training labels and the difficulty of annotating, dealing with adverse driving conditions such as nighttime has posed a huge challenge to the perception system of autonomous vehicles. Therefore, adapting knowledge from a labelled daytime domain to an unlabelled nighttime domain has been widely researched. In addition to labelled daytime datasets, existing nighttime datasets usually provide nighttime images with corresponding daytime reference images captured at nearby locations for reference. The key challenge is to minimize the performance gap between the two domains. In this paper, we propose LoopDA for domain adaptive nighttime semantic segmentation. It consists of self-loops that result in reconstructing the input data using predicted semantic maps, by rendering them into the encoded features. In a warm-up training stage, the self-loops comprise of an inner-loop and an outer-loop, which are responsible for intra-domain refinement and inter-domain alignment, respectively. To reduce the impact of day-night pose shifts, in the later self-training stage, we propose a co-teaching pipeline that involves an offline pseudo-supervision signal and an online reference-guided signal `DNA' (Day-Night Agreement), bringing substantial benefits to enhance nighttime segmentation. Our model outperforms prior methods on Dark Zurich and Nighttime Driving datasets for semantic segmentation. Code and pretrained models are available at https://github.com/fy-vision/LoopDA., Comment: Accepted to WACV2023

Human Activity Recognition (HAR) has been employed in a wide range of applications, e.g. self-driving cars, where safety and lives are at stake. Recently, the robustness of skeleton-based HAR methods have been questioned due to their vulnerability to adversarial attacks. However, the proposed attacks require the full-knowledge of the attacked classifier, which is overly restrictive. In this paper, we show such threats indeed exist, even when the attacker only has access to the input/output of the model. To this end, we propose the very first black-box adversarial attack approach in skeleton-based HAR called BASAR. BASAR explores the interplay between the classification boundary and the natural motion manifold. To our best knowledge, this is the first time data manifold is introduced in adversarial attacks on time series. Via BASAR, we find on-manifold adversarial samples are extremely deceitful and rather common in skeletal motions, in contrast to the common belief that adversarial samples only exist off-manifold. Through exhaustive evaluation, we show that BASAR can deliver successful attacks across classifiers, datasets, and attack modes. By attack, BASAR helps identify the potential causes of the model vulnerability and provides insights on possible improvements. Finally, to mitigate the newly identified threat, we propose a new adversarial training approach by leveraging the sophisticated distributions of on/off-manifold adversarial samples, called mixed manifold-based adversarial training (MMAT). MMAT can successfully help defend against adversarial attacks without compromising classification accuracy., Comment: Accepted in Pattern Recognition. arXiv admin note: substantial text overlap with arXiv:2103.05266

As one of the detection targets of all gravitational wave detectors at present, stochastic gravitational wave background (SGWB) provides us an important way to understand the evolution of our universe. In this paper, we explore the feasibility of detecting the SGWB generated by the loops, which arose throughout the cosmological evolution of the cosmic string network, by individual space detectors (e.g.~LISA, TAIJI) and joint space detectors (LISA-TAIJI). For joint detectors, we choose three different configurations of TAIJI (e.g.~TAIJIm, TAIJIp, TAIJIc) to form the LISA-TAIJI networks. And we investigate the performance of them to detect the SGWB. Though comparing the power-law sensitivity (PLS) curves of individual space detectors and joint detectors with energy density spectrum of SGWB. We find that LISA-TAIJIc has the best sensitivity for detecting the SGWB from cosmic string loops and is promising to further constrains the tension of cosmic sting $G\mu=O(10^{-17})$.

For years, researchers have been devoted to generalizable object perception and manipulation, where cross-category generalizability is highly desired yet underexplored. In this work, we propose to learn such cross-category skills via Generalizable and Actionable Parts (GAParts). By identifying and defining 9 GAPart classes (lids, handles, etc.) in 27 object categories, we construct a large-scale part-centric interactive dataset, GAPartNet, where we provide rich, part-level annotations (semantics, poses) for 8,489 part instances on 1,166 objects. Based on GAPartNet, we investigate three cross-category tasks: part segmentation, part pose estimation, and part-based object manipulation. Given the significant domain gaps between seen and unseen object categories, we propose a robust 3D segmentation method from the perspective of domain generalization by integrating adversarial learning techniques. Our method outperforms all existing methods by a large margin, no matter on seen or unseen categories. Furthermore, with part segmentation and pose estimation results, we leverage the GAPart pose definition to design part-based manipulation heuristics that can generalize well to unseen object categories in both the simulator and the real world. Our dataset, code, and demos are available on our project page., Comment: To appear in CVPR 2023 (Highlight)

Pixel-wise prediction with deep neural network has become an effective paradigm for salient object detection (SOD) and achieved remarkable performance. However, very few SOD models are robust against adversarial attacks which are visually imperceptible for human visual attention. The previous work robust saliency (ROSA) shuffles the pre-segmented superpixels and then refines the coarse saliency map by the densely connected conditional random field (CRF). Different from ROSA that relies on various pre- and post-processings, this paper proposes a light-weight Learnable Noise (LeNo) to defend adversarial attacks for SOD models. LeNo preserves accuracy of SOD models on both adversarial and clean images, as well as inference speed. In general, LeNo consists of a simple shallow noise and noise estimation that embedded in the encoder and decoder of arbitrary SOD networks respectively. Inspired by the center prior of human visual attention mechanism, we initialize the shallow noise with a cross-shaped gaussian distribution for better defense against adversarial attacks. Instead of adding additional network components for post-processing, the proposed noise estimation modifies only one channel of the decoder. With the deeply-supervised noise-decoupled training on state-of-the-art RGB and RGB-D SOD networks, LeNo outperforms previous works not only on adversarial images but also on clean images, which contributes stronger robustness for SOD. Our code is available at https://github.com/ssecv/LeNo., Comment: 8 pages, 6 figures, accepted by AAAI 2023

In this work, we tackle 6-DoF grasp detection for transparent and specular objects, which is an important yet challenging problem in vision-based robotic systems, due to the failure of depth cameras in sensing their geometry. We, for the first time, propose a multiview RGB-based 6-DoF grasp detection network, GraspNeRF, that leverages the generalizable neural radiance field (NeRF) to achieve material-agnostic object grasping in clutter. Compared to the existing NeRF-based 3-DoF grasp detection methods that rely on densely captured input images and time-consuming per-scene optimization, our system can perform zero-shot NeRF construction with sparse RGB inputs and reliably detect 6-DoF grasps, both in real-time. The proposed framework jointly learns generalizable NeRF and grasp detection in an end-to-end manner, optimizing the scene representation construction for the grasping. For training data, we generate a large-scale photorealistic domain-randomized synthetic dataset of grasping in cluttered tabletop scenes that enables direct transfer to the real world. Our extensive experiments in synthetic and real-world environments demonstrate that our method significantly outperforms all the baselines in all the experiments while remaining in real-time. Project page can be found at https://pku-epic.github.io/GraspNeRF, Comment: IEEE International Conference on Robotics and Automation (ICRA), 2023

The quality of cardiac magnetic resonance (CMR) imaging is susceptible to respiratory motion artifacts. The model robustness of automated segmentation techniques in face of real-world respiratory motion artifacts is unclear. This manuscript describes the design of extreme cardiac MRI analysis challenge under respiratory motion (CMRxMotion Challenge). The challenge aims to establish a public benchmark dataset to assess the effects of respiratory motion on image quality and examine the robustness of segmentation models. The challenge recruited 40 healthy volunteers to perform different breath-hold behaviors during one imaging visit, obtaining paired cine imaging with artifacts. Radiologists assessed the image quality and annotated the level of respiratory motion artifacts. For those images with diagnostic quality, radiologists further segmented the left ventricle, left ventricle myocardium and right ventricle. The images of training set (20 volunteers) along with the annotations are released to the challenge participants, to develop an automated image quality assessment model (Task 1) and an automated segmentation model (Task 2). The images of validation set (5 volunteers) are released to the challenge participants but the annotations are withheld for online evaluation of submitted predictions. Both the images and annotations of the test set (15 volunteers) were withheld and only used for offline evaluation of submitted containerized dockers. The image quality assessment task is quantitatively evaluated by the Cohen's kappa statistics and the segmentation task is evaluated by the Dice scores and Hausdorff distances., Comment: Summary of CMRxMotion Challenge Design

When manipulating an object to accomplish complex tasks, humans rely on both vision and touch to keep track of the object's 6D pose. However, most existing object pose tracking systems in robotics rely exclusively on visual signals, which hinder a robot's ability to manipulate objects effectively. To address this limitation, we introduce TEG-Track, a tactile-enhanced 6D pose tracking system that can track previously unseen objects held in hand. From consecutive tactile signals, TEG-Track optimizes object velocities from marker flows when slippage does not occur, or regresses velocities using a slippage estimation network when slippage is detected. The estimated object velocities are integrated into a geometric-kinematic optimization scheme to enhance existing visual pose trackers. To evaluate our method and to facilitate future research, we construct a real-world dataset for visual-tactile in-hand object pose tracking. Experimental results demonstrate that TEG-Track consistently enhances state-of-the-art generalizable 6D pose trackers in synthetic and real-world scenarios. Our code and dataset are available at https://github.com/leolyliu/TEG-Track.

Robotic dexterous grasping is the first step to enable human-like dexterous object manipulation and thus a crucial robotic technology. However, dexterous grasping is much more under-explored than object grasping with parallel grippers, partially due to the lack of a large-scale dataset. In this work, we present a large-scale robotic dexterous grasp dataset, DexGraspNet, generated by our proposed highly efficient synthesis method that can be generally applied to any dexterous hand. Our method leverages a deeply accelerated differentiable force closure estimator and thus can efficiently and robustly synthesize stable and diverse grasps on a large scale. We choose ShadowHand and generate 1.32 million grasps for 5355 objects, covering more than 133 object categories and containing more than 200 diverse grasps for each object instance, with all grasps having been validated by the Isaac Gym simulator. Compared to the previous dataset from Liu et al. generated by GraspIt!, our dataset has not only more objects and grasps, but also higher diversity and quality. Via performing cross-dataset experiments, we show that training several algorithms of dexterous grasp synthesis on our dataset significantly outperforms training on the previous one. To access our data and code, including code for human and Allegro grasp synthesis, please visit our project page: https://pku-epic.github.io/DexGraspNet/.

As a decentralized training approach, federated learning enables multiple organizations to jointly train a model without exposing their private data. This work investigates vertical federated learning (VFL) to address scenarios where collaborating organizations have the same set of users but with different features, and only one party holds the labels. While VFL shows good performance, practitioners often face uncertainty when preparing non-transparent, internal/external features and samples for the VFL training phase. Moreover, to balance the prediction accuracy and the resource consumption of model inference, practitioners require to know which subset of prediction instances is genuinely needed to invoke the VFL model for inference. To this end, we co-design the VFL modeling process by proposing an interactive real-time visualization system, VFLens, to help practitioners with feature engineering, sample selection, and inference. A usage scenario, a quantitative experiment, and expert feedback suggest that VFLens helps practitioners boost VFL efficiency at a lower cost with sufficient confidence., Comment: The Tenth International Symposium of Chinese CHI (Chinese CHI 2022), October 22--23, 2022, Guangzhou, China and Online, China

We present a novel traffic trajectory editing method which uses spatio-temporal keyframes to control vehicles during the simulation to generate desired traffic trajectories. By taking self-motivation, path following and collision avoidance into account, the proposed force-based traffic simulation framework updates vehicle's motions in both the Frenet coordinates and the Cartesian coordinates. With the way-points from users, lane-level navigation can be generated by reference path planning. With a given keyframe, the coarse-to-fine optimization is proposed to efficiently generate the plausible trajectory which can satisfy the spatio-temporal constraints. At first, a directed state-time graph constructed along the reference path is used to search for a coarse-grained trajectory by mapping the keyframe as the goal. Then, using the information extracted from the coarse trajectory as initialization, adjoint-based optimization is applied to generate a finer trajectory with smooth motions based on our force-based simulation. We validate our method with extensive experiments.

In this work, we tackle the challenging task of jointly tracking hand object pose and reconstructing their shapes from depth point cloud sequences in the wild, given the initial poses at frame 0. We for the first time propose a point cloud based hand joint tracking network, HandTrackNet, to estimate the inter-frame hand joint motion. Our HandTrackNet proposes a novel hand pose canonicalization module to ease the tracking task, yielding accurate and robust hand joint tracking. Our pipeline then reconstructs the full hand via converting the predicted hand joints into a template-based parametric hand model MANO. For object tracking, we devise a simple yet effective module that estimates the object SDF from the first frame and performs optimization-based tracking. Finally, a joint optimization step is adopted to perform joint hand and object reasoning, which alleviates the occlusion-induced ambiguity and further refines the hand pose. During training, the whole pipeline only sees purely synthetic data, which are synthesized with sufficient variations and by depth simulation for the ease of generalization. The whole pipeline is pertinent to the generalization gaps and thus directly transferable to real in-the-wild data. We evaluate our method on two real hand object interaction datasets, e.g. HO3D and DexYCB, without any finetuning. Our experiments demonstrate that the proposed method significantly outperforms the previous state-of-the-art depth-based hand and object pose estimation and tracking methods, running at a frame rate of 9 FPS.

We present the results of the first Machine Learning Gravitational-Wave Search Mock Data Challenge (MLGWSC-1). For this challenge, participating groups had to identify gravitational-wave signals from binary black hole mergers of increasing complexity and duration embedded in progressively more realistic noise. The final of the 4 provided datasets contained real noise from the O3a observing run and signals up to a duration of 20 seconds with the inclusion of precession effects and higher order modes. We present the average sensitivity distance and runtime for the 6 entered algorithms derived from 1 month of test data unknown to the participants prior to submission. Of these, 4 are machine learning algorithms. We find that the best machine learning based algorithms are able to achieve up to 95% of the sensitive distance of matched-filtering based production analyses for simulated Gaussian noise at a false-alarm rate (FAR) of one per month. In contrast, for real noise, the leading machine learning search achieved 70%. For higher FARs the differences in sensitive distance shrink to the point where select machine learning submissions outperform traditional search algorithms at FARs $\geq 200$ per month on some datasets. Our results show that current machine learning search algorithms may already be sensitive enough in limited parameter regions to be useful for some production settings. To improve the state-of-the-art, machine learning algorithms need to reduce the false-alarm rates at which they are capable of detecting signals and extend their validity to regions of parameter space where modeled searches are computationally expensive to run. Based on our findings we compile a list of research areas that we believe are the most important to elevate machine learning searches to an invaluable tool in gravitational-wave signal detection., Comment: 25 pages, 6 figures, 4 tables, additional material available at https://github.com/gwastro/ml-mock-data-challenge-1

Single-view point cloud completion aims to recover the full geometry of an object based on only limited observation, which is extremely hard due to the data sparsity and occlusion. The core challenge is to generate plausible geometries to fill the unobserved part of the object based on a partial scan, which is under-constrained and suffers from a huge solution space. Inspired by the classic shadow volume technique in computer graphics, we propose a new method to reduce the solution space effectively. Our method considers the camera a light source that casts rays toward the object. Such light rays build a reasonably constrained but sufficiently expressive basis for completion. The completion process is then formulated as a point displacement optimization problem. Points are initialized at the partial scan and then moved to their goal locations with two types of movements for each point: directional movements along the light rays and constrained local movement for shape refinement. We design neural networks to predict the ideal point movements to get the completion results. We demonstrate that our method is accurate, robust, and generalizable through exhaustive evaluation and comparison. Moreover, it outperforms state-of-the-art methods qualitatively and quantitatively on MVP datasets., Comment: SIGGRAPH Aisa 2022 Conference Paper

We propose a novel method for generating high-resolution videos of talking-heads from speech audio and a single 'identity' image. Our method is based on a convolutional neural network model that incorporates a pre-trained StyleGAN generator. We model each frame as a point in the latent space of StyleGAN so that a video corresponds to a trajectory through the latent space. Training the network is in two stages. The first stage is to model trajectories in the latent space conditioned on speech utterances. To do this, we use an existing encoder to invert the generator, mapping from each video frame into the latent space. We train a recurrent neural network to map from speech utterances to displacements in the latent space of the image generator. These displacements are relative to the back-projection into the latent space of an identity image chosen from the individuals depicted in the training dataset. In the second stage, we improve the visual quality of the generated videos by tuning the image generator on a single image or a short video of any chosen identity. We evaluate our model on standard measures (PSNR, SSIM, FID and LMD) and show that it significantly outperforms recent state-of-the-art methods on one of two commonly used datasets and gives comparable performance on the other. Finally, we report on ablation experiments that validate the components of the model. The code and videos from experiments can be found at https://mohammedalghamdi.github.io/talking-heads-acm-mm, Comment: Accepted at ACM Multimedia 2022. The Project webpage can found at https://mohammedalghamdi.github.io/talking-heads-acm-mm