We present a novel method for reconstructing personalized 3D human avatars with realistic animation from only a few images. Due to the large variations in body shapes, poses, and cloth types, existing methods mostly require hours of per-subject optimization during inference, which limits their practical applications. In contrast, we learn a universal prior from over a thousand clothed humans to achieve instant feedforward generation and zero-shot generalization. Specifically, instead of rigging the avatar with shared skinning weights, we jointly infer personalized avatar shape, skinning weights, and pose-dependent deformations, which effectively improves overall geometric fidelity and reduces deformation artifacts. Moreover, to normalize pose variations and resolve coupled ambiguity between canonical shapes and skinning weights, we design a 3D canonicalization process to produce pixel-aligned initial conditions, which helps to reconstruct fine-grained geometric details. We then propose a multi-frame feature aggregation to robustly reduce artifacts introduced in canonicalization and fuse a plausible avatar preserving person-specific identities. Finally, we train the model in an end-to-end framework on a large-scale capture dataset, which contains diverse human subjects paired with high-quality 3D scans. Extensive experiments show that our method generates more authentic reconstruction and animation than state-of-the- arts, and can be directly generalized to inputs from casually taken phone photos. Project page and code is available at https://github.com/rongakowang/FRESA.

Existing photorealistic relightable hand models require extensive identity-specific observations in different views, poses, and illuminations, and face challenges in generalizing to natural illuminations and novel identities. To bridge this gap, we present URHand, the first universal relightable hand model that generalizes across viewpoints, poses, illuminations, and identities. Our model allows few-shot personalization using images captured with a mobile phone, and is ready to be photorealistically rendered under novel illuminations. To simplify the personalization process while retaining photorealism, we build a powerful universal relightable prior based on neural relighting from multi-view images of hands captured in a light stage with hundreds of identities. The key challenge is scaling the cross-identity training while maintaining personalized fidelity and sharp details without compromising generalization under natural illuminations. To this end, we propose a spatially varying linear lighting model as the neural renderer that takes physics-inspired shading as input feature. By removing non-linear activations and bias, our specifically designed lighting model explicitly keeps the linearity of light transport. This enables single-stage training from light-stage data while generalizing to real-time rendering under arbitrary continuous illuminations across diverse identities. In addition, we introduce the joint learning of a physically based model and our neural relighting model, which further improves fidelity and generalization. Extensive experiments show that our approach achieves superior performance over existing methods in terms of both quality and generalizability. We also demonstrate quick personalization of URHand from a short phone scan of an unseen identity.

The authentic 3D hand avatar with every identifiable information, such as hand shapes and textures, is necessary for immersive experiences in AR/VR. In this paper, we present a universal hand model (UHM), which 1) can universally represent high-fidelity 3D hand meshes of arbitrary identities (IDs) and 2) can be adapted to each person with a short phone scan for the authentic hand avatar. For effective universal hand modeling, we perform tracking and modeling at the same time, while previous 3D hand models perform them separately. The conventional separate pipeline suffers from the accumulated errors from the tracking stage, which cannot be recovered in the modeling stage. On the other hand, ours does not suffer from the accumulated errors while having a much more concise overall pipeline. We additionally introduce a novel image matching loss function to address a skin sliding during the tracking and modeling, while existing works have not focused on it much. Finally, using learned priors from our UHM, we effectively adapt our UHM to each person's short phone scan for the authentic hand avatar.

To build photorealistic avatars that users can embody, human modelling must be complete (cover the full body), driveable (able to reproduce the current motion and appearance from the user), and generalizable (i.e., easily adaptable to novel identities).Towards these goals, paired captures, that is, captures of the same subject obtained from systems of diverse quality and availability, are crucial.However, paired captures are rarely available to researchers outside of dedicated industrial labs: Codec Avatar Studio is our proposal to close this gap.Towards generalization and driveability, we introduce a dataset of 256 subjects captured in two modalities: high resolution multi-view scans of their heads, and video from the internal cameras of a headset.Towards completeness, we introduce a dataset of 4 subjects captured in eight modalities: high quality relightable multi-view captures of heads and hands, full body multi-view captures with minimal and regular clothes, and corresponding head, hands and body phone captures.Together with our data, we also provide code and pre-trained models for different state-of-the-art human generation models.Our datasets and code are available at https://github.com/facebookresearch/ava-256 and https://github.com/facebookresearch/goliath.

We present the first neural relighting approach for rendering high-fidelity personalized hands that can be animated in real-time under novel illumination. Our approach adopts a teacher-student framework, where the teacher learns appearance under a single point light from images captured in a light-stage, allowing us to synthesize hands in arbitrary illuminations but with heavy compute. Using images rendered by the teacher model as training data, an efficient student model directly predicts appearance under natural illuminations in real-time. To achieve generalization, we condition the student model with physics-inspired illumination features such as visibility, diffuse shading, and specular reflections computed on a coarse proxy geometry, maintaining a small computational overhead. Our key insight is that these features have strong correlation with subsequent global light transport effects, which proves sufficient as conditioning data for the neural relighting network. Moreover, in contrast to bottleneck illumination conditioning, these features are spatially aligned based on underlying geometry, leading to better generalization to unseen illuminations and poses. In our experiments, we demonstrate the efficacy of our illumination feature representations, outperforming baseline approaches. We also show that our approach can photorealistically relight two interacting hands at real-time speeds. https://sh8.io/#/relightable_hands

The two-hand interaction is one of the most challenging signals to analyze due to the self-similarity, complicated articulations, and occlusions of hands. Although several datasets have been proposed for the two-hand interaction analysis, all of them do not achieve 1) diverse and realistic image appearances and 2) diverse and large-scale groundtruth (GT) 3D poses at the same time. In this work, we propose Re:InterHand, a dataset of relighted 3D interacting hands that achieve the two goals. To this end, we employ a state-of-the-art hand relighting network with our accurately tracked two-hand 3D poses. We compare our Re:InterHand with existing 3D interacting hands datasets and show the benefit of it. Our Re:InterHand is available in https://mks0601.github.io/ReInterHand/

We study the use of raw ultrasound waveforms, often referred to as the "Radio Frequency" (RF) data, for the semantic segmentation of ultrasound scans to carry out dense and diagnostic labeling. We present W-Net, a novel Convolution Neural Network (CNN) framework that employs the raw ultrasound waveforms in addition to the grey ultrasound image to semantically segment and label tissues for anatomical, pathological, or other diagnostic purposes. To the best of our knowledge, this is also the first deep-learning or CNN approach for segmentation that analyzes ultrasound raw RF data along with the grey image. We chose subcutaneous tissue (SubQ) segmentation as our initial clinical goal for dense segmentation since it has diverse intermixed tissues, is challenging to segment, and is an underrepresented research area. SubQ potential applications include plastic surgery, adipose stem-cell harvesting, lymphatic monitoring, and possibly detection/treatment of certain types of tumors. Unlike prior work, we seek to label every pixel in the image, without the use of a background class. A custom dataset consisting of hand-labeled images by an expert clinician and trainees are used for the experimentation, currently labeled into the following categories: skin, fat, fat fascia/stroma, muscle, and muscle fascia. We compared W-Net and attention variant of W-Net (AW-Net) with U-Net and Attention U-Net (AU-Net). Our novel W-Net's RF-Waveform encoding architecture outperformed regular U-Net and AU-Net, achieving the best mIoU accuracy (averaged across all tissue classes). We study the impact of RF data on dense labeling of the SubQ region, which is followed by the analyses of the generalization capability of the networks to patients and analysis on the SubQ tissue classes, determining that fascia tissues, especially muscle fascia in particular, are the most difficult anatomic class to recognize for both humans and AI algorithms. We present diagnostic semantic segmentation, which is semantic segmentation carried out for the purposes of direct diagnostic pixel labeling, and apply it to breast tumor detection task on a publicly available dataset to segment pixels into malignant tumor, benign tumor, and background tissue class. Using the segmented image we diagnose the patient by classifying the breast lesion as either benign or malignant. We demonstrate the diagnostic capability of RF data with the use of W-Net, which achieves the best segmentation scores across all classes.

This study in preterm infants was designed to characterize the prognostic potential of several features of heart rate variability (HRV), respiration, and (infant) motion for the predictive monitoring of late-onset sepsis (LOS). In a neonatal intensive care setting, the cardiorespiratory waveforms of infants with blood-culture positive LOS were analyzed to characterize the prognostic potential of 22 features for discriminating control from sepsis-state, using the Naïve Bayes algorithm. Historical data of the subjects acquired from a period sufficiently before the clinical suspicion of LOS was used as control state, whereas data from the 24 h preceding the clinical suspicion of LOS were used as sepsis state (test data). The overall prognostic potential of all features was quantified at three-hourly intervals for the period corresponding to test data by calculating the area under the receiver operating characteristics curve. For the 49 infants studied, features of HRV, respiration, and movement showed characteristic changes in the hours leading up to the clinical suspicion of sepsis, namely, an increased propensity toward pathological heart rate decelerations, increased respiratory instability, and a decrease in spontaneous infant activity, i.e., lethargy. While features characterizing HRV and respiration can be used to probe the state of the autonomic nervous system, those characterizing movement probe the state of the motor system-dysregulation of both reflects an increased likelihood of sepsis. By using readily interpretable features derived from cardiorespiratory monitoring, opportunities for pre-emptively identifying and treating LOS can be developed.

Blood pressure (BP) and heart rate (HR) are often captured in conjunction with weight in telemonitoring systems, but the additional prognostic potential of daily measurements of BP and HR in providing information on upcoming hospitalizations for worsening heart failure (HFH) have not been explored thoroughly. We retrospectively analyzed 267 daily home-telemonitored heart failure (HF) subjects. We extracted those episodes of HFHs that had sufficient data entries in the days leading up to hospitalization and tested the prognostic potential of 48 trend features based on weight, systolic BP, diastolic BP, pulse pressure (PP), and HR with a Naïve Bayesian model. The single best-performing trend feature-with a cross-validated estimate of 0.64 for the area under the curve (AUC) with a standard deviation (SD) of 0.01-is based on a 2-day weight trend. The best multivariate feature set (cross-validated [Formula: see text], [Formula: see text]) comprises of 2-day trend features based on weight, systolic BP, and HR. There were large variations in the weight trends preceding hospitalizations and weight change alone had a modest predictive ability. Readily interpretable features capturing trends in BP and HR provided additional prognostic information and can be used for improving classification.