We present MoGe, a powerful model for recovering 3D geometry from monocular open-domain images. Given a single image, our model directly predicts a 3D point map of the captured scene with an affine-invariant representation, which is agnostic to true global scale and shift. This new representation precludes ambiguous supervision in training and facilitates effective geometry learning. Furthermore, we propose a set of novel global and local geometry supervision techniques that empower the model to learn high-quality geometry. These include a robust, optimal, and efficient point cloud alignment solver for accurate global shape learning, and a multi-scale local geometry loss promoting precise local geometry supervision. We train our model on a large, mixed dataset and demonstrate its strong generalizability and high accuracy. In our comprehensive evaluation on diverse unseen datasets, our model significantly outperforms state-of-the-art methods across all tasks, including monocular estimation of 3D point map, depth map, and camera field of view.

We introduce a novel 3D generation method for versatile and high-quality 3D asset creation. The cornerstone is a unified Structured LATent (SLat) representation which allows decoding to different output formats, such as Radiance Fields, 3D Gaussians, and meshes. This is achieved by integrating a sparsely-populated 3D grid with dense multiview visual features extracted from a powerful vision foundation model, comprehensively capturing both structural (geometry) and textural (appearance) information while maintaining flexibility during decoding.We employ rectified flow transformers tailored for SLat as our 3D generation models and train models with up to 2 billion parameters on a large 3D asset dataset of 500K diverse objects. Our model generates high-quality results with text or image conditions, significantly surpassing existing methods, including recent ones at similar scales. We showcase flexible output format selection and local 3D editing capabilities which were not offered by previous models. Project Page: trellis3d.github.io.

In this paper, we present a novel framework for video-to-4D generation that creates high-quality dynamic 3D content from single video inputs. Direct 4D diffusion modeling is extremely challenging due to costly data construction and the high-dimensional nature of jointly representing 3D shape, appearance, and motion. We address these challenges by introducing a Direct 4DMesh-to-GS Variation Field VAE that directly encodes canonical Gaussian Splats (GS) and their temporal variations from 3D animation data without per-instance fitting, and compresses high-dimensional animations into a compact latent space. Building upon this efficient representation, we train a Gaussian Variation Field diffusion model with temporal-aware Diffusion Transformer conditioned on input videos and canonical GS. Trained on carefully-curated animatable 3D objects from the Objaverse dataset, our model demonstrates superior generation quality compared to existing methods. It also exhibits remarkable generalization to in-the-wild video inputs despite being trained exclusively on synthetic data, paving the way for generating high-quality animated 3D content. Project page: https://gvfdiffusion.github.io/.

Robotic manipulation in cluttered environments presents significant challenges, particularly when the clutter includes thin, deformable objects like cables, which complicate perception and decision-making processes. In the context of datacenters, the automation of networking tasks often involves the manipulation of optical transceivers within densely packed cable configurations. Such environments are characterized by an abundance of delicate, overlapping, and intersecting cables, leading to frequent occlusions. This paper introduces an innovative system designed for the manipulation of optical transceivers in environments cluttered by cables. Our integrated approach combines advanced 3D scene understanding with a heuristic-based pushing policy to effectively manipulate optical transceivers amidst clutter. The system's perception component utilizes image segmentation and 3D reconstruction to accurately model the transceivers and surrounding cables. Meanwhile, the planning aspect employs a search algorithm with task-specific heuristics, to navigate the gripper, displace obstructing cables, and safely achieve a precise pre-grasp position in front of the target transceiver. We have conducted extensive evaluations of our methodology in both simulated and real-world settings, demonstrating its high success rates, robustness, and proficiency in addressing the unique challenges posed by cable-occluded environments within datacenters.

Dropout [1], a standard regularization technique in Large Language Models, incurs extra computational overhead, particularly due to its repeated application during the training process. To address this issue, we propose Accelerated Dropout, an algorithm that revolutionizes the traditional dropout method by employing a bitmask, rather than a float mask, to significantly expedite the training process. In the theoretical proof, we demonstrate that accelerated dropout is exponentially convergent to the probability of retaining neurons. Our extensive experimental analysis, conducted on 13 benchmark datasets and 9 deep learning models, confirms that accelerated dropout outperforms traditional dropout in terms of training efficiency and generalization performance. The experimental results indicate that, compared to torch dropout, our accelerated dropout achieves 8x speedup on a single dropout operator. With the use of accelerated dropout, the average training speed of each model increased by 1.0624x. The training speed improvements ranged from a 1.046x increase for ProphetNet on the CNN/DailyMail dataset to a 1.077x increase for ViT-B on the ImageNet dataset.

The rate-distortion (RD) theory is one of the key concepts in information theory, providing theoretical limits for compression performance and guiding the source coding design, with both theoretical and practical significance. The Blahut-Arimoto (BA) algorithm, as a classical algorithm to compute RD functions, encounters computational challenges when applied to high-dimensional scenarios. In recent years, many neural methods have attempted to compute high-dimensional RD problems from the perspective of implicit generative models. Nevertheless, these approaches often neglect the reconstruction of the optimal conditional distribution or rely on unreasonable prior assumptions. In face of these issues, we propose an innovative energy-based modeling framework that leverages the connection between the RD dual form and the free energy in statistical physics, achieving effective reconstruction of the optimal conditional distribution. The proposed algorithm requires training only a single neural network and circumvents the challenge of computing the normalization factor in energy-based models using the Markov chain Monte Carlo (MCMC) sampling. Experimental results demonstrate the significant effectiveness of the proposed algorithm in estimating high-dimensional RD functions and reconstructing the optimal conditional distribution.

We propose MoGe-2, an advanced open-domain geometry estimation model that recovers a metric-scale 3D point map of a scene from a single image. Our method builds upon the recent monocular geometry estimation approach, MoGe, which predicts affine-invariant point maps with unknown scales. We explore effective strategies to extend MoGe for metric geometry prediction without compromising the relative geometry accuracy provided by the affine-invariant point representation. Additionally, we discover that noise and errors in real data diminish fine-grained detail in the predicted geometry. We address this by developing a data refinement approach that filters and completes real data using sharp synthetic labels, significantly enhancing the granularity of the reconstructed geometry while maintaining the overall accuracy. We train our model on a large corpus of mixed datasets and conducted comprehensive evaluations, demonstrating its superior performance in achieving accurate relative geometry, precise metric scale, and fine-grained detail recovery -- capabilities that no previous methods have simultaneously achieved.

We propose VASA-3D, an audio-driven, single-shot 3D head avatar generator. This research tackles two major challenges: capturing the subtle expression details present in real human faces, and reconstructing an intricate 3D head avatar from a single portrait image. To accurately model expression details, VASA-3D leverages the motion latent of VASA-1, a method that yields exceptional realism and vividness in 2D talking heads. A critical element of our work is translating this motion latent to 3D, which is accomplished by devising a 3D head model that is conditioned on the motion latent. Customization of this model to a single image is achieved through an optimization framework that employs numerous video frames of the reference head synthesized from the input image. The optimization takes various training losses robust to artifacts and limited pose coverage in the generated training data. Our experiment shows that VASA-3D produces realistic 3D talking heads that cannot be achieved by prior art, and it supports the online generation of 512x512 free-viewpoint videos at up to 75 FPS, facilitating more immersive engagements with lifelike 3D avatars.

Audio-driven 3D facial animation is crucial for enhancing the metaverse's realism, immersion, and interactivity. While most existing methods focus on generating highly realistic and lively 2D talking head videos by leveraging extensive 2D video datasets these approaches work in pixel space and are not easily adaptable to 3D environments. We present VASA-Rig, which has achieved a significant advancement in the realism of lip-audio synchronization, facial dynamics, and head movements. In particular, we introduce a novel rig parameter-based emotional talking face dataset and propose the Latents2Rig model, which facilitates the transformation of 2D facial animations into 3D. Unlike mesh-based models, VASA-Rig outputs rig parameters, instantiated in this paper as 174 Metahuman rig parameters, making it more suitable for integration into industry-standard pipelines. Extensive experimental results demonstrate that our approach significantly outperforms existing state-of-the-art methods in terms of both realism and accuracy.

The information bottleneck (IB) method is a technique designed to extract meaningful information related to one random variable from another random variable, and has found extensive applications in machine learning problems. In this paper, neural network based estimation of the IB problem solution is studied, through the lens of a novel formulation of the IB problem. Via exploiting the inherent structure of the IB functional and leveraging the mapping approach, the proposed formulation of the IB problem involves only a single variable to be optimized, and subsequently is readily amenable to data-driven estimators based on neural networks. A theoretical analysis is conducted to guarantee that the neural estimator asymptotically solves the IB problem, and the numerical experiments on both synthetic and MNIST datasets demonstrate the effectiveness of the neural estimator.

We introduce VASA, a framework for generating lifelike talking faces with appealing visual affective skills (VAS) given a single static image and a speech audio clip. Our premiere model, VASA-1, is capable of not only generating lip movements that are exquisitely synchronized with the audio, but also producing a large spectrum of facial nuances and natural head motions that contribute to the perception of authenticity and liveliness. The core innovations include a diffusion-based holistic facial dynamics and head movement generation model that works in a face latent space, and the development of such an expressive and disentangled face latent space using videos. Through extensive experiments including evaluation on a set of new metrics, we show that our method significantly outperforms previous methods along various dimensions comprehensively. Our method delivers high video quality with realistic facial and head dynamics and also supports the online generation of 512$\times$512 videos at up to 40 FPS with negligible starting latency. It paves the way for real-time engagements with lifelike avatars that emulate human conversational behaviors.

Previous animatable 3D-aware GANs for human generation have primarily focused on either the human head or full body. However, head-only videos are relatively uncommon in real life, and full body generation typically does not deal with facial expression control and still has challenges in generating high-quality results. Towards applicable video avatars, we present an animatable 3D-aware GAN that generates portrait images with controllable facial expression, head pose, and shoulder movements. It is a generative model trained on unstructured 2D image collections without using 3D or video data. For the new task, we base our method on the generative radiance manifold representation and equip it with learnable facial and head-shoulder deformations. A dual-camera rendering and adversarial learning scheme is proposed to improve the quality of the generated faces, which is critical for portrait images. A pose deformation processing network is developed to generate plausible deformations for challenging regions such as long hair. Experiments show that our method, trained on unstructured 2D images, can generate diverse and high-quality 3D portraits with desired control over different properties.

Recent research advance has significantly improved the visual real-ism of immersive 3D video communication. In this work we present a method to further enhance this immersive experience by adding the hand touch capability (“remote hand clapping”). In our system, each meeting participant sits in front of a large screen with haptic feedback. The local participant can reach his hand out to the screen and perform hand clapping with the remote participant as if the two participants were only separated by a virtual glass. A key challenge in emulating the remote hand touch is the realistic rendering of the participant's hand and arm as the hand touches the screen. When the hand is very close to the screen, the RGBD data required for realistic rendering is no longer available. To tackle this challenge, we present a dual representation of the user's hand. Our dual representation not only preserves the high-quality rendering usually found in recent image-based rendering systems but also allows the hand to reach to the screen. This is possible because the dual representation includes both an image-based model and a 3D geometry-based model, with the latter driven by a hand skeleton tracked by a side view camera. In addition, the dual representation provides a distance-based fusion of the image-based and 3D geometry-based models as the hand moves closer to the screen. The result is that the image-based and 3D geometry-based models mutually enhance each other, leading to realistic and seamless rendering. Our experiments demonstrate that our method provides consistent hand contact experience between remote users and improves the immersive experience of 3D video communication.

In this paper, we present a learning-based approach for recovering the 3D geometry of human head from a single portrait image. Our method is learned in an unsupervised manner without any ground-truth 3D data. We represent the head geometry with a parametric 3D face model together with a depth map for other head regions including hair and ear. A two-step geometry learning scheme is proposed to learn 3D head reconstruction from in-the-wild face images, where we first learn face shape on single images using self-reconstruction and then learn hair and ear geometry using pairs of images in a stereo-matching fashion. The second step is based on the output of the first to not only improve the accuracy but also ensure the consistency of overall head geometry. We evaluate the accuracy of our method both in 3D and with pose manipulation tasks on 2D images. We alter pose based on the recovered geometry and apply a refinement network trained with adversarial learning to ameliorate the reprojected images and translate them to the real image domain. Extensive evaluations and comparison with previous methods show that our new method can produce high-fidelity 3D head geometry and head pose manipulation results.