We present MotionBERT, a unified pretraining framework, to tackle different sub-tasks of human motion analysis including 3D pose estimation, skeleton-based action recognition, and mesh recovery. The proposed framework is capable of utilizing all kinds of human motion data resources, including motion capture data and in-the-wild videos. During pretraining, the pretext task requires the motion encoder to recover the underlying 3D motion from noisy partial 2D observations. The pretrained motion representation thus acquires geometric, kinematic, and physical knowledge about human motion and therefore can be easily transferred to multiple downstream tasks. We implement the motion encoder with a novel Dual-stream Spatio-temporal Transformer (DSTformer) neural network. It could capture long-range spatio-temporal relationships among the skeletal joints comprehensively and adaptively, exemplified by the lowest 3D pose estimation error so far when trained from scratch. More importantly, the proposed framework achieves state-of-the-art performance on all three downstream tasks by simply finetuning the pretrained motion encoder with 1-2 linear layers, which demonstrates the versatility of the learned motion representations.
Realistic generative face video synthesis has long been a pursuit in both computer vision and graphics community. However, existing face video generation methods tend to produce low-quality frames with drifted facial identities and unnatural movements. To tackle these challenges, we propose a principled framework named StyleFaceV, which produces high-fidelity identity-preserving face videos with vivid movements. Our core insight is to decompose appearance and pose information and recompose them in the latent space of StyleGAN3 to produce stable and dynamic results. Specifically, StyleGAN3 provides strong priors for high-fidelity facial image generation, but the latent space is intrinsically entangled. By carefully examining its latent properties, we propose our decomposition and recomposition designs which allow for the disentangled combination of facial appearance and movements. Moreover, a temporal-dependent model is built upon the decomposed latent features, and samples reasonable sequences of motions that are capable of generating realistic and temporally coherent face videos. Particularly, our pipeline is trained with a joint training strategy on both static images and high-quality video data, which is of higher data efficiency. Extensive experiments demonstrate that our framework achieves state-of-the-art face video generation results both qualitatively and quantitatively. Notably, StyleFaceV is capable of generating realistic $1024\times1024$ face videos even without high-resolution training videos.
Large-scale datasets have played indispensable roles in the recent success of face generation/editing and significantly facilitated the advances of emerging research fields. However, the academic community still lacks a video dataset with diverse facial attribute annotations, which is crucial for the research on face-related videos. In this work, we propose a large-scale, high-quality, and diverse video dataset with rich facial attribute annotations, named the High-Quality Celebrity Video Dataset (CelebV-HQ). CelebV-HQ contains 35,666 video clips with the resolution of 512x512 at least, involving 15,653 identities. All clips are labeled manually with 83 facial attributes, covering appearance, action, and emotion. We conduct a comprehensive analysis in terms of age, ethnicity, brightness stability, motion smoothness, head pose diversity, and data quality to demonstrate the diversity and temporal coherence of CelebV-HQ. Besides, its versatility and potential are validated on two representative tasks, i.e., unconditional video generation and video facial attribute editing. Furthermore, we envision the future potential of CelebV-HQ, as well as the new opportunities and challenges it would bring to related research directions. Data, code, and models are publicly available. Project page: https://celebv-hq.github.io.
Video-to-Video synthesis (Vid2Vid) has achieved remarkable results in generating a photo-realistic video from a sequence of semantic maps. However, this pipeline suffers from high computational cost and long inference latency, which largely depends on two essential factors: 1) network architecture parameters, 2) sequential data stream. Recently, the parameters of image-based generative models have been significantly compressed via more efficient network architectures. Nevertheless, existing methods mainly focus on slimming network architectures and ignore the size of the sequential data stream. Moreover, due to the lack of temporal coherence, image-based compression is not sufficient for the compression of the video task. In this paper, we present a spatial-temporal compression framework, \textbf{Fast-Vid2Vid}, which focuses on data aspects of generative models. It makes the first attempt at time dimension to reduce computational resources and accelerate inference. Specifically, we compress the input data stream spatially and reduce the temporal redundancy. After the proposed spatial-temporal knowledge distillation, our model can synthesize key-frames using the low-resolution data stream. Finally, Fast-Vid2Vid interpolates intermediate frames by motion compensation with slight latency. On standard benchmarks, Fast-Vid2Vid achieves around real-time performance as 20 FPS and saves around 8x computational cost on a single V100 GPU.
Generic event boundary detection (GEBD) is an important yet challenging task in video understanding, which aims at detecting the moments where humans naturally perceive event boundaries. In this paper, we present a local context modeling and global boundary decoding approach for GEBD task. Local context modeling sub-network is proposed to perceive diverse patterns of generic event boundaries, and it generates powerful video representations and reliable boundary confidence. Based on them, global boundary decoding sub-network is exploited to decode event boundaries from a global view. Our proposed method achieves 85.13% F1-score on Kinetics-GEBD testing set, which achieves a more than 22% F1-score boost compared to the baseline method. The code is available at https://github.com/JackyTown/GEBD_Challenge_CVPR2022.
Generating high-quality and diverse human images is an important yet challenging task in vision and graphics. However, existing generative models often fall short under the high diversity of clothing shapes and textures. Furthermore, the generation process is even desired to be intuitively controllable for layman users. In this work, we present a text-driven controllable framework, Text2Human, for a high-quality and diverse human generation. We synthesize full-body human images starting from a given human pose with two dedicated steps. 1) With some texts describing the shapes of clothes, the given human pose is first translated to a human parsing map. 2) The final human image is then generated by providing the system with more attributes about the textures of clothes. Specifically, to model the diversity of clothing textures, we build a hierarchical texture-aware codebook that stores multi-scale neural representations for each type of texture. The codebook at the coarse level includes the structural representations of textures, while the codebook at the fine level focuses on the details of textures. To make use of the learned hierarchical codebook to synthesize desired images, a diffusion-based transformer sampler with mixture of experts is firstly employed to sample indices from the coarsest level of the codebook, which then is used to predict the indices of the codebook at finer levels. The predicted indices at different levels are translated to human images by the decoder learned accompanied with hierarchical codebooks. The use of mixture-of-experts allows for the generated image conditioned on the fine-grained text input. The prediction for finer level indices refines the quality of clothing textures. Extensive quantitative and qualitative evaluations demonstrate that our proposed framework can generate more diverse and realistic human images compared to state-of-the-art methods.
Although significant progress has been made to audio-driven talking face generation, existing methods either neglect facial emotion or cannot be applied to arbitrary subjects. In this paper, we propose the Emotion-Aware Motion Model (EAMM) to generate one-shot emotional talking faces by involving an emotion source video. Specifically, we first propose an Audio2Facial-Dynamics module, which renders talking faces from audio-driven unsupervised zero- and first-order key-points motion. Then through exploring the motion model's properties, we further propose an Implicit Emotion Displacement Learner to represent emotion-related facial dynamics as linearly additive displacements to the previously acquired motion representations. Comprehensive experiments demonstrate that by incorporating the results from both modules, our method can generate satisfactory talking face results on arbitrary subjects with realistic emotion patterns.
This paper focuses on the weakly-supervised audio-visual video parsing task, which aims to recognize all events belonging to each modality and localize their temporal boundaries. This task is challenging because only overall labels indicating the video events are provided for training. However, an event might be labeled but not appear in one of the modalities, which results in a modality-specific noisy label problem. Motivated by two observations that networks tend to learn clean samples first and that a labeled event would appear in at least one modality, we propose a training strategy to identify and remove modality-specific noisy labels dynamically. Specifically, we sort the losses of all instances within a mini-batch individually in each modality, then select noisy samples according to relationships between intra-modal and inter-modal losses. Besides, we also propose a simple but valid noise ratio estimation method by calculating the proportion of instances whose confidence is below a preset threshold. Our method makes large improvements over the previous state of the arts (e.g., from 60.0% to 63.8% in segment-level visual metric), which demonstrates the effectiveness of our approach.
Unconditional human image generation is an important task in vision and graphics, which enables various applications in the creative industry. Existing studies in this field mainly focus on "network engineering" such as designing new components and objective functions. This work takes a data-centric perspective and investigates multiple critical aspects in "data engineering", which we believe would complement the current practice. To facilitate a comprehensive study, we collect and annotate a large-scale human image dataset with over 230K samples capturing diverse poses and textures. Equipped with this large dataset, we rigorously investigate three essential factors in data engineering for StyleGAN-based human generation, namely data size, data distribution, and data alignment. Extensive experiments reveal several valuable observations w.r.t. these aspects: 1) Large-scale data, more than 40K images, are needed to train a high-fidelity unconditional human generation model with vanilla StyleGAN. 2) A balanced training set helps improve the generation quality with rare face poses compared to the long-tailed counterpart, whereas simply balancing the clothing texture distribution does not effectively bring an improvement. 3) Human GAN models with body centers for alignment outperform models trained using face centers or pelvis points as alignment anchors. In addition, a model zoo and human editing applications are demonstrated to facilitate future research in the community.
This work targets at using a general deep learning framework to synthesize free-viewpoint images of arbitrary human performers, only requiring a sparse number of camera views as inputs and skirting per-case fine-tuning. The large variation of geometry and appearance, caused by articulated body poses, shapes and clothing types, are the key bottlenecks of this task. To overcome these challenges, we present a simple yet powerful framework, named Generalizable Neural Performer (GNR), that learns a generalizable and robust neural body representation over various geometry and appearance. Specifically, we compress the light fields for novel view human rendering as conditional implicit neural radiance fields from both geometry and appearance aspects. We first introduce an Implicit Geometric Body Embedding strategy to enhance the robustness based on both parametric 3D human body model and multi-view images hints. We further propose a Screen-Space Occlusion-Aware Appearance Blending technique to preserve the high-quality appearance, through interpolating source view appearance to the radiance fields with a relax but approximate geometric guidance. To evaluate our method, we present our ongoing effort of constructing a dataset with remarkable complexity and diversity. The dataset GeneBody-1.0, includes over 360M frames of 370 subjects under multi-view cameras capturing, performing a large variety of pose actions, along with diverse body shapes, clothing, accessories and hairdos. Experiments on GeneBody-1.0 and ZJU-Mocap show better robustness of our methods than recent state-of-the-art generalizable methods among all cross-dataset, unseen subjects and unseen poses settings. We also demonstrate the competitiveness of our model compared with cutting-edge case-specific ones. Dataset, code and model will be made publicly available.