Wanhua Li

Shuai Shen, Wanhua Li, Zheng Zhu et al. · tsinghua

Talking head synthesis is an emerging technology with wide applications in film dubbing, virtual avatars and online education. Recent NeRF-based methods generate more natural talking videos, as they better capture the 3D structural information of faces. However, a specific model needs to be trained for each identity with a large dataset. In this paper, we propose Dynamic Facial Radiance Fields (DFRF) for few-shot talking head synthesis, which can rapidly generalize to an unseen identity with few training data. Different from the existing NeRF-based methods which directly encode the 3D geometry and appearance of a specific person into the network, our DFRF conditions face radiance field on 2D appearance images to learn the face prior. Thus the facial radiance field can be flexibly adjusted to the new identity with few reference images. Additionally, for better modeling of the facial deformations, we propose a differentiable face warping module conditioned on audio signals to deform all reference images to the query space. Extensive experiments show that with only tens of seconds of training clip available, our proposed DFRF can synthesize natural and high-quality audio-driven talking head videos for novel identities with only 40k iterations. We highly recommend readers view our supplementary video for intuitive comparisons. Code is available in https://sstzal.github.io/DFRF/.

Devaansh Gupta, Siddhant Kharbanda, Jiawei Zhou et al. · microsoft-research, tsinghua

There has been a growing interest in developing multimodal machine translation (MMT) systems that enhance neural machine translation (NMT) with visual knowledge. This problem setup involves using images as auxiliary information during training, and more recently, eliminating their use during inference. Towards this end, previous works face a challenge in training powerful MMT models from scratch due to the scarcity of annotated multilingual vision-language data, especially for low-resource languages. Simultaneously, there has been an influx of multilingual pre-trained models for NMT and multimodal pre-trained models for vision-language tasks, primarily in English, which have shown exceptional generalisation ability. However, these are not directly applicable to MMT since they do not provide aligned multimodal multilingual features for generative tasks. To alleviate this issue, instead of designing complex modules for MMT, we propose CLIPTrans, which simply adapts the independently pre-trained multimodal M-CLIP and the multilingual mBART. In order to align their embedding spaces, mBART is conditioned on the M-CLIP features by a prefix sequence generated through a lightweight mapping network. We train this in a two-stage pipeline which warms up the model with image captioning before the actual translation task. Through experiments, we demonstrate the merits of this framework and consequently push forward the state-of-the-art across standard benchmarks by an average of +2.67 BLEU. The code can be found at www.github.com/devaansh100/CLIPTrans.

Wanhua Li, Xiaoke Huang, Zheng Zhu et al. · tsinghua

This paper presents a language-powered paradigm for ordinal regression. Existing methods usually treat each rank as a category and employ a set of weights to learn these concepts. These methods are easy to overfit and usually attain unsatisfactory performance as the learned concepts are mainly derived from the training set. Recent large pre-trained vision-language models like CLIP have shown impressive performance on various visual tasks. In this paper, we propose to learn the rank concepts from the rich semantic CLIP latent space. Specifically, we reformulate this task as an image-language matching problem with a contrastive objective, which regards labels as text and obtains a language prototype from a text encoder for each rank. While prompt engineering for CLIP is extremely time-consuming, we propose OrdinalCLIP, a differentiable prompting method for adapting CLIP for ordinal regression. OrdinalCLIP consists of learnable context tokens and learnable rank embeddings; The learnable rank embeddings are constructed by explicitly modeling numerical continuity, resulting in well-ordered, compact language prototypes in the CLIP space. Once learned, we can only save the language prototypes and discard the huge language model, resulting in zero additional computational overhead compared with the linear head counterpart. Experimental results show that our paradigm achieves competitive performance in general ordinal regression tasks, and gains improvements in few-shot and distribution shift settings for age estimation. The code is available at https://github.com/xk-huang/OrdinalCLIP.

Wanhua Li, Zhexuan Cao, Jianjiang Feng et al. · tsinghua

Objects are usually associated with multiple attributes, and these attributes often exhibit high correlations. Modeling complex relationships between attributes poses a great challenge for multi-attribute learning. This paper proposes a simple yet generic framework named Label2Label to exploit the complex attribute correlations. Label2Label is the first attempt for multi-attribute prediction from the perspective of language modeling. Specifically, it treats each attribute label as a "word" describing the sample. As each sample is annotated with multiple attribute labels, these "words" will naturally form an unordered but meaningful "sentence", which depicts the semantic information of the corresponding sample. Inspired by the remarkable success of pre-training language models in NLP, Label2Label introduces an image-conditioned masked language model, which randomly masks some of the "word" tokens from the label "sentence" and aims to recover them based on the masked "sentence" and the context conveyed by image features. Our intuition is that the instance-wise attribute relations are well grasped if the neural net can infer the missing attributes based on the context and the remaining attribute hints. Label2Label is conceptually simple and empirically powerful. Without incorporating task-specific prior knowledge and highly specialized network designs, our approach achieves state-of-the-art results on three different multi-attribute learning tasks, compared to highly customized domain-specific methods. Code is available at https://github.com/Li-Wanhua/Label2Label.

Shuai Shen, Wenliang Zhao, Zibin Meng et al. · tsinghua

Talking head synthesis is a promising approach for the video production industry. Recently, a lot of effort has been devoted in this research area to improve the generation quality or enhance the model generalization. However, there are few works able to address both issues simultaneously, which is essential for practical applications. To this end, in this paper, we turn attention to the emerging powerful Latent Diffusion Models, and model the Talking head generation as an audio-driven temporally coherent denoising process (DiffTalk). More specifically, instead of employing audio signals as the single driving factor, we investigate the control mechanism of the talking face, and incorporate reference face images and landmarks as conditions for personality-aware generalized synthesis. In this way, the proposed DiffTalk is capable of producing high-quality talking head videos in synchronization with the source audio, and more importantly, it can be naturally generalized across different identities without any further fine-tuning. Additionally, our DiffTalk can be gracefully tailored for higher-resolution synthesis with negligible extra computational cost. Extensive experiments show that the proposed DiffTalk efficiently synthesizes high-fidelity audio-driven talking head videos for generalized novel identities. For more video results, please refer to \url{https://sstzal.github.io/DiffTalk/}.

Wanhua Li, Jiwen Lu, Abudukelimu Wuerkaixi et al. · tsinghua

Different people age in different ways. Learning a personalized age estimator for each person is a promising direction for age estimation given that it better models the personalization of aging processes. However, most existing personalized methods suffer from the lack of large-scale datasets due to the high-level requirements: identity labels and enough samples for each person to form a long-term aging pattern. In this paper, we aim to learn personalized age estimators without the above requirements and propose a meta-learning method named MetaAge for age estimation. Unlike most existing personalized methods that learn the parameters of a personalized estimator for each person in the training set, our method learns the mapping from identity information to age estimator parameters. Specifically, we introduce a personalized estimator meta-learner, which takes identity features as the input and outputs the parameters of customized estimators. In this way, our method learns the meta knowledge without the above requirements and seamlessly transfers the learned meta knowledge to the test set, which enables us to leverage the existing large-scale age datasets without any additional annotations. Extensive experimental results on three benchmark datasets including MORPH II, ChaLearn LAP 2015 and ChaLearn LAP 2016 databases demonstrate that our MetaAge significantly boosts the performance of existing personalized methods and outperforms the state-of-the-art approaches.

Chenyu Wu, Wanhua Li, Zhu-Tian Chen et al.

Reconstructing dynamic 3D scenes from monocular videos is a fundamental yet highly challenging task, as real-world motions often involve both long-term smooth transformations and short-term complex deformations. Existing methods either struggle to maintain temporal consistency or fail to capture high-frequency dynamics due to limited motion modeling capacity. In this work, we present Rigid-aware 4D Gaussian Splatting (RiGS), which simultaneously captures motions across multiple temporal scales. Specifically, RiGS introduces three types of Gaussian primitives: static, rigid, and transient, which represent static backgrounds, long-term low-frequency motions, and short-term high-frequency dynamics, respectively. An object-wise dynamic mask is proposed to aggregate long-range spatiotemporal motion information and guide the decomposition of static and dynamic regions. To jointly model motion across scales, rigid Gaussians are allowed to transition into transient Gaussians based on their temporal duration, and both are optimized under scene flow guidance, providing dense 3D motion supervision. Extensive experiments demonstrate that RiGS achieves state-of-the-art performance on novel view synthesis benchmarks. Code is available at \hyperlink{https://github.com/ladvu/RiGS}{https://github.com/ladvu/RiGS}.

Yilong Liu, Wanhua Li, Chen Zhu-Tian et al.

We present LangFlash, a feed-forward framework for 3D Language Gaussian Splatting that reconstructs 3D scenes parameterized by Gaussian primitives enriched with language-aligned semantic features from sparse unposed multi-view images. Unlike optimization-based 3D methods, LangFlash directly predicts the geometry and semantics in a single forward pass, enabling low-latency 3D reconstruction and language-consistent scene understanding. To support large-scale training, we enriched the RealEstate10k dataset with coherent and dense semantic information for 3D semantic supervision. Furthermore, we propose a sparse semantic encoding scheme that combines a global semantic dictionary with locally varying per-primitive weights, preserving high-level linguistic information, while reducing representation complexity. Experimental results show that LangFlash achieves superior novel view synthesis and semantic consistency compared with previous methods. This study establishes a new paradigm for pose-free, language-grounded 3D scene reconstruction, advancing generalizable 3D vision and multimodal scene understanding. Demo is available at https://liylo.github.io/langflash.github.io/.

Zhiyi Shi, Junsik Kim, Wanhua Li et al.

Multi-modal pre-trained models efficiently extract and fuse features from different modalities with low memory requirements for fine-tuning. Despite this efficiency, their application in disease diagnosis is under-explored. A significant challenge is the frequent occurrence of missing modalities, which impairs performance. Additionally, fine-tuning the entire pre-trained model demands substantial computational resources. To address these issues, we introduce Modality-aware Low-Rank Adaptation (MoRA), a computationally efficient method. MoRA projects each input to a low intrinsic dimension but uses different modality-aware up-projections for modality-specific adaptation in cases of missing modalities. Practically, MoRA integrates into the first block of the model, significantly improving performance when a modality is missing. It requires minimal computational resources, with less than 1.6% of the trainable parameters needed compared to training the entire model. Experimental results show that MoRA outperforms existing techniques in disease diagnosis, demonstrating superior performance, robustness, and training efficiency.

Ye Liu, Jixuan He, Wanhua Li et al.

Video temporal grounding (VTG) is a fine-grained video understanding problem that aims to ground relevant clips in untrimmed videos given natural language queries. Most existing VTG models are built upon frame-wise final-layer CLIP features, aided by additional temporal backbones (e.g., SlowFast) with sophisticated temporal reasoning mechanisms. In this work, we claim that CLIP itself already shows great potential for fine-grained spatial-temporal modeling, as each layer offers distinct yet useful information under different granularity levels. Motivated by this, we propose Reversed Recurrent Tuning ($R^2$-Tuning), a parameter- and memory-efficient transfer learning framework for video temporal grounding. Our method learns a lightweight $R^2$ Block containing only 1.5% of the total parameters to perform progressive spatial-temporal modeling. Starting from the last layer of CLIP, $R^2$ Block recurrently aggregates spatial features from earlier layers, then refines temporal correlation conditioning on the given query, resulting in a coarse-to-fine scheme. $R^2$-Tuning achieves state-of-the-art performance across three VTG tasks (i.e., moment retrieval, highlight detection, and video summarization) on six public benchmarks (i.e., QVHighlights, Charades-STA, Ego4D-NLQ, TACoS, YouTube Highlights, and TVSum) even without the additional backbone, demonstrating the significance and effectiveness of the proposed scheme. Our code is available at https://github.com/yeliudev/R2-Tuning.

Yichen Wu, Hongming Piao, Long-Kai Huang et al.

Continual Learning (CL) with foundation models has recently emerged as a promising paradigm to exploit abundant knowledge acquired during pre-training for tackling sequential tasks. However, existing prompt-based and Low-Rank Adaptation-based (LoRA-based) methods often require expanding a prompt/LoRA pool or retaining samples of previous tasks, which poses significant scalability challenges as the number of tasks grows. To address these limitations, we propose Scalable Decoupled LoRA (SD-LoRA) for class incremental learning, which continually separates the learning of the magnitude and direction of LoRA components without rehearsal. Our empirical and theoretical analysis reveals that SD-LoRA tends to follow a low-loss trajectory and converges to an overlapping low-loss region for all learned tasks, resulting in an excellent stability-plasticity trade-off. Building upon these insights, we introduce two variants of SD-LoRA with further improved parameter efficiency. All parameters of SD-LoRAs can be end-to-end optimized for CL objectives. Meanwhile, they support efficient inference by allowing direct evaluation with the finally trained model, obviating the need for component selection. Extensive experiments across multiple CL benchmarks and foundation models consistently validate the effectiveness of SD-LoRA. The code is available at https://github.com/WuYichen-97/SD-Lora-CL.

Wanhua Li, Zibin Meng, Jiawei Zhou et al.

Social relation reasoning aims to identify relation categories such as friends, spouses, and colleagues from images. While current methods adopt the paradigm of training a dedicated network end-to-end using labeled image data, they are limited in terms of generalizability and interpretability. To address these issues, we first present a simple yet well-crafted framework named {\name}, which combines the perception capability of Vision Foundation Models (VFMs) and the reasoning capability of Large Language Models (LLMs) within a modular framework, providing a strong baseline for social relation recognition. Specifically, we instruct VFMs to translate image content into a textual social story, and then utilize LLMs for text-based reasoning. {\name} introduces systematic design principles to adapt VFMs and LLMs separately and bridge their gaps. Without additional model training, it achieves competitive zero-shot results on two databases while offering interpretable answers, as LLMs can generate language-based explanations for the decisions. The manual prompt design process for LLMs at the reasoning phase is tedious and an automated prompt optimization method is desired. As we essentially convert a visual classification task into a generative task of LLMs, automatic prompt optimization encounters a unique long prompt optimization issue. To address this issue, we further propose the Greedy Segment Prompt Optimization (GSPO), which performs a greedy search by utilizing gradient information at the segment level. Experimental results show that GSPO significantly improves performance, and our method also generalizes to different image styles. The code is available at https://github.com/Mengzibin/SocialGPT.

Jixuan He, Wanhua Li, Ye Liu et al.

As a common image editing operation, image composition involves integrating foreground objects into background scenes. In this paper, we expand the application of the concept of Affordance from human-centered image composition tasks to a more general object-scene composition framework, addressing the complex interplay between foreground objects and background scenes. Following the principle of Affordance, we define the affordance-aware object insertion task, which aims to seamlessly insert any object into any scene with various position prompts. To address the limited data issue and incorporate this task, we constructed the SAM-FB dataset, which contains over 3 million examples across more than 3,000 object categories. Furthermore, we propose the Mask-Aware Dual Diffusion (MADD) model, which utilizes a dual-stream architecture to simultaneously denoise the RGB image and the insertion mask. By explicitly modeling the insertion mask in the diffusion process, MADD effectively facilitates the notion of affordance. Extensive experimental results show that our method outperforms the state-of-the-art methods and exhibits strong generalization performance on in-the-wild images. Please refer to our code on https://github.com/KaKituken/affordance-aware-any.

Yifan Liu, Fangneng Zhan, Wanhua Li et al.

Estimating robot pose from a monocular RGB image is a challenge in robotics and computer vision. Existing methods typically build networks on top of 2D visual backbones and depend heavily on labeled data for training, which is often scarce in real-world scenarios, causing a sim-to-real gap. Moreover, these approaches reduce the 3D-based problem to 2D domain, neglecting the 3D priors. To address these, we propose Robot Topological Alignment Graph (RoboTAG), which incorporates a 3D branch to inject 3D priors while enabling co-evolution of the 2D and 3D representations, alleviating the reliance on labels. Specifically, the RoboTAG consists of a 3D branch and a 2D branch, where nodes represent the states of the camera and robot system, and edges capture the dependencies between these variables or denote alignments between them. Closed loops are then defined in the graph, on which a consistency supervision across branches can be applied. This design allows us to utilize in-the-wild images as training data without annotations. Experimental results demonstrate that our method is effective across robot types, highlighting its potential to alleviate the data bottleneck in robotics.

Minghan Qin, Wanhua Li, Jiawei Zhou et al. · tsinghua

Humans live in a 3D world and commonly use natural language to interact with a 3D scene. Modeling a 3D language field to support open-ended language queries in 3D has gained increasing attention recently. This paper introduces LangSplat, which constructs a 3D language field that enables precise and efficient open-vocabulary querying within 3D spaces. Unlike existing methods that ground CLIP language embeddings in a NeRF model, LangSplat advances the field by utilizing a collection of 3D Gaussians, each encoding language features distilled from CLIP, to represent the language field. By employing a tile-based splatting technique for rendering language features, we circumvent the costly rendering process inherent in NeRF. Instead of directly learning CLIP embeddings, LangSplat first trains a scene-wise language autoencoder and then learns language features on the scene-specific latent space, thereby alleviating substantial memory demands imposed by explicit modeling. Existing methods struggle with imprecise and vague 3D language fields, which fail to discern clear boundaries between objects. We delve into this issue and propose to learn hierarchical semantics using SAM, thereby eliminating the need for extensively querying the language field across various scales and the regularization of DINO features. Extensive experimental results show that LangSplat significantly outperforms the previous state-of-the-art method LERF by a large margin. Notably, LangSplat is extremely efficient, achieving a 199 $\times$ speedup compared to LERF at the resolution of 1440 $\times$ 1080. We strongly recommend readers to check out our video results at https://langsplat.github.io/

Tong Ding, Wanhua Li, Zhongqi Miao et al.

Prompt learning has proven effective in adapting vision language models for downstream tasks. However, existing methods usually append learnable prompt tokens solely with the category names to obtain textual features, which fails to fully leverage the rich context indicated in the category name. To address this issue, we propose the Tree of Attributes Prompt learning (TAP), which first instructs LLMs to generate a tree of attributes with a "concept - attribute - description" structure for each category, and then learn the hierarchy with vision and text prompt tokens. Unlike existing methods that merely augment category names with a set of unstructured descriptions, our approach essentially distills structured knowledge graphs associated with class names from LLMs. Furthermore, our approach introduces text and vision prompts designed to explicitly learn the corresponding visual attributes, effectively serving as domain experts. Additionally, the general and diverse descriptions generated based on the class names may be wrong or absent in the specific given images. To address this misalignment, we further introduce a vision-conditional pooling module to extract instance-specific text features. Extensive experimental results demonstrate that our approach outperforms state-of-the-art methods on the zero-shot base-to-novel generalization, cross-dataset transfer, as well as few-shot classification across 11 diverse datasets. Code is available at https://github.com/HHenryD/TAP.

Juanxi Tian, Fengyuan Liu, Jiaming Han et al.

Aligning multimodal generative models with human preferences demands reward signals that respect the compositional, multi-dimensional structure of human judgment. Prevailing RLHF approaches reduce this structure to scalar or pairwise labels, collapsing nuanced preferences into opaque parametric proxies and exposing vulnerabilities to reward hacking. While recent Rubrics-as-Reward (RaR) methods attempt to recover this structure through explicit criteria, generating rubrics that are simultaneously reliable, scalable, and data-efficient remains an open problem. We introduce Auto-Rubric as Reward (ARR), a framework that reframes reward modeling from implicit weight optimization to explicit, criteria-based decomposition. Before any pairwise comparison, ARR externalizes a VLM's internalized preference knowledge as prompt-specific rubrics, translating holistic intent into independently verifiable quality dimensions. This conversion of implicit preference structure into inspectable, interpretable constraints substantially suppresses evaluation biases including positional bias, enabling both zero-shot deployment and few-shot conditioning on minimal supervision. To extend these gains into generative training, we propose Rubric Policy Optimization (RPO), which distills ARR's structured multi-dimensional evaluation into a robust binary reward, replacing opaque scalar regression with rubric-conditioned preference decisions that stabilize policy gradients. On text-to-image generation and image editing benchmarks, ARR-RPO outperforms pairwise reward models and VLM judges, demonstrating that explicitly externalizing implicit preference knowledge into structured rubrics achieves more reliable, data-efficient multimodal alignment, revealing that the bottleneck is the absence of a factorized interface, not a deficit of knowledge.

Wanhua Li, Xiaoke Huang, Jiwen Lu et al.

Uncertainty is the only certainty there is. Modeling data uncertainty is essential for regression, especially in unconstrained settings. Traditionally the direct regression formulation is considered and the uncertainty is modeled by modifying the output space to a certain family of probabilistic distributions. On the other hand, classification based regression and ranking based solutions are more popular in practice while the direct regression methods suffer from the limited performance. How to model the uncertainty within the present-day technologies for regression remains an open issue. In this paper, we propose to learn probabilistic ordinal embeddings which represent each data as a multivariate Gaussian distribution rather than a deterministic point in the latent space. An ordinal distribution constraint is proposed to exploit the ordinal nature of regression. Our probabilistic ordinal embeddings can be integrated into popular regression approaches and empower them with the ability of uncertainty estimation. Experimental results show that our approach achieves competitive performance. Code is available at https://github.com/Li-Wanhua/POEs.

Shuai Shen, Wanhua Li, Zheng Zhu et al.

Face clustering is a promising method for annotating unlabeled face images. Recent supervised approaches have boosted the face clustering accuracy greatly, however their performance is still far from satisfactory. These methods can be roughly divided into global-based and local-based ones. Global-based methods suffer from the limitation of training data scale, while local-based ones are difficult to grasp the whole graph structure information and usually take a long time for inference. Previous approaches fail to tackle these two challenges simultaneously. To address the dilemma of large-scale training and efficient inference, we propose the STructure-AwaRe Face Clustering (STAR-FC) method. Specifically, we design a structure-preserved subgraph sampling strategy to explore the power of large-scale training data, which can increase the training data scale from ${10^{5}}$ to ${10^{7}}$. During inference, the STAR-FC performs efficient full-graph clustering with two steps: graph parsing and graph refinement. And the concept of node intimacy is introduced in the second step to mine the local structural information. The STAR-FC gets 91.97 pairwise F-score on partial MS1M within 310s which surpasses the state-of-the-arts. Furthermore, we are the first to train on very large-scale graph with 20M nodes, and achieve superior inference results on 12M testing data. Overall, as a simple and effective method, the proposed STAR-FC provides a strong baseline for large-scale face clustering. Code is available at \url{https://sstzal.github.io/STAR-FC/}.

Wanhua Li, Renping Zhou, Jiawei Zhou et al.

Learning 4D language fields to enable time-sensitive, open-ended language queries in dynamic scenes is essential for many real-world applications. While LangSplat successfully grounds CLIP features into 3D Gaussian representations, achieving precision and efficiency in 3D static scenes, it lacks the ability to handle dynamic 4D fields as CLIP, designed for static image-text tasks, cannot capture temporal dynamics in videos. Real-world environments are inherently dynamic, with object semantics evolving over time. Building a precise 4D language field necessitates obtaining pixel-aligned, object-wise video features, which current vision models struggle to achieve. To address these challenges, we propose 4D LangSplat, which learns 4D language fields to handle time-agnostic or time-sensitive open-vocabulary queries in dynamic scenes efficiently. 4D LangSplat bypasses learning the language field from vision features and instead learns directly from text generated from object-wise video captions via Multimodal Large Language Models (MLLMs). Specifically, we propose a multimodal object-wise video prompting method, consisting of visual and text prompts that guide MLLMs to generate detailed, temporally consistent, high-quality captions for objects throughout a video. These captions are encoded using a Large Language Model into high-quality sentence embeddings, which then serve as pixel-aligned, object-specific feature supervision, facilitating open-vocabulary text queries through shared embedding spaces. Recognizing that objects in 4D scenes exhibit smooth transitions across states, we further propose a status deformable network to model these continuous changes over time effectively. Our results across multiple benchmarks demonstrate that 4D LangSplat attains precise and efficient results for both time-sensitive and time-agnostic open-vocabulary queries.

Jixuan Fan, Wanhua Li, Yifei Han et al.

3D Gaussian Splatting has demonstrated notable success in large-scale scene reconstruction, but challenges persist due to high training memory consumption and storage overhead. Hybrid representations that integrate implicit and explicit features offer a way to mitigate these limitations. However, when applied in parallelized block-wise training, two critical issues arise since reconstruction accuracy deteriorates due to reduced data diversity when training each block independently, and parallel training restricts the number of divided blocks to the available number of GPUs. To address these issues, we propose Momentum-GS, a novel approach that leverages momentum-based self-distillation to promote consistency and accuracy across the blocks while decoupling the number of blocks from the physical GPU count. Our method maintains a teacher Gaussian decoder updated with momentum, ensuring a stable reference during training. This teacher provides each block with global guidance in a self-distillation manner, promoting spatial consistency in reconstruction. To further ensure consistency across the blocks, we incorporate block weighting, dynamically adjusting each block's weight according to its reconstruction accuracy. Extensive experiments on large-scale scenes show that our method consistently outperforms existing techniques, achieving a 12.8% improvement in LPIPS over CityGaussian with much fewer divided blocks and establishing a new state of the art. Project page: https://jixuan-fan.github.io/Momentum-GS_Page/

Kento Nishi, Junsik Kim, Wanhua Li et al.

Multi-task learning has become increasingly popular in the machine learning field, but its practicality is hindered by the need for large, labeled datasets. Most multi-task learning methods depend on fully labeled datasets wherein each input example is accompanied by ground-truth labels for all target tasks. Unfortunately, curating such datasets can be prohibitively expensive and impractical, especially for dense prediction tasks which require per-pixel labels for each image. With this in mind, we propose Joint-Task Regularization (JTR), an intuitive technique which leverages cross-task relations to simultaneously regularize all tasks in a single joint-task latent space to improve learning when data is not fully labeled for all tasks. JTR stands out from existing approaches in that it regularizes all tasks jointly rather than separately in pairs -- therefore, it achieves linear complexity relative to the number of tasks while previous methods scale quadratically. To demonstrate the validity of our approach, we extensively benchmark our method across a wide variety of partially labeled scenarios based on NYU-v2, Cityscapes, and Taskonomy.

Wanhua Li, Yujie Zhao, Minghan Qin et al.

In this paper, we introduce LangSplatV2, which achieves high-dimensional feature splatting at 476.2 FPS and 3D open-vocabulary text querying at 384.6 FPS for high-resolution images, providing a 42 $\times$ speedup and a 47 $\times$ boost over LangSplat respectively, along with improved query accuracy. LangSplat employs Gaussian Splatting to embed 2D CLIP language features into 3D, significantly enhancing speed and learning a precise 3D language field with SAM semantics. Such advancements in 3D language fields are crucial for applications that require language interaction within complex scenes. However, LangSplat does not yet achieve real-time inference performance (8.2 FPS), even with advanced A100 GPUs, severely limiting its broader application. In this paper, we first conduct a detailed time analysis of LangSplat, identifying the heavyweight decoder as the primary speed bottleneck. Our solution, LangSplatV2 assumes that each Gaussian acts as a sparse code within a global dictionary, leading to the learning of a 3D sparse coefficient field that entirely eliminates the need for a heavyweight decoder. By leveraging this sparsity, we further propose an efficient sparse coefficient splatting method with CUDA optimization, rendering high-dimensional feature maps at high quality while incurring only the time cost of splatting an ultra-low-dimensional feature. Our experimental results demonstrate that LangSplatV2 not only achieves better or competitive query accuracy but is also significantly faster. Codes and demos are available at our project page: https://langsplat-v2.github.io.

Junsik Kim, Zhiyi Shi, Davin Jeong et al.

In clinical In-Vitro Fertilization (IVF), identifying the most viable embryo for transfer is important to increasing the likelihood of a successful pregnancy. Traditionally, this process involves embryologists manually assessing embryos' static morphological features at specific intervals using light microscopy. This manual evaluation is not only time-intensive and costly, due to the need for expert analysis, but also inherently subjective, leading to variability in the selection process. To address these challenges, we develop a multimodal model that leverages both time-lapse video data and Electronic Health Records (EHRs) to predict embryo viability. One of the primary challenges of our research is to effectively combine time-lapse video and EHR data, owing to their inherent differences in modality. We comprehensively analyze our multimodal model with various modality inputs and integration approaches. Our approach will enable fast and automated embryo viability predictions in scale for clinical IVF.

Karly Hou, Wanhua Li, Hanspeter Pfister

Recently, Gaussian Splatting methods have emerged as a desirable substitute for prior Radiance Field methods for novel-view synthesis of scenes captured with multi-view images or videos. In this work, we propose a novel extension to 4D Gaussian Splatting for dynamic scenes. Drawing on ideas from residual learning, we hierarchically decompose the dynamic scene into a "video-segment-frame" structure, with segments dynamically adjusted by optical flow. Then, instead of directly predicting the time-dependent signals, we model the signal as the sum of video-constant values, segment-constant values, and frame-specific residuals, as inspired by the success of residual learning. This approach allows more flexible models that adapt to highly variable scenes. We demonstrate state-of-the-art visual quality and real-time rendering on several established datasets, with the greatest improvements on complex scenes with large movements, occlusions, and fine details, where current methods degrade most.

Shuai Shen, Wanhua Li, Yunpeng Zhang et al.

Talking head synthesis has emerged as a prominent research topic in computer graphics and multimedia, yet most existing methods often struggle to strike a balance between generation quality and computational efficiency, particularly under real-time constraints. In this paper, we propose a novel framework that integrates Gaussian Splatting with a structured Audio Factorization Plane (Audio-Plane) to enable high-quality, audio-synchronized, and real-time talking head generation. For modeling a dynamic talking head, a 4D volume representation, which consists of three axes in 3D space and one temporal axis aligned with audio progression, is typically required. However, directly storing and processing a dense 4D grid is impractical due to the high memory and computation cost, and lack of scalability for longer durations. We address this challenge by decomposing the 4D volume representation into a set of audio-independent spatial planes and audio-dependent planes, forming a compact and interpretable representation for talking head modeling that we refer to as the Audio-Plane. This factorized design allows for efficient and fine-grained audio-aware spatial encoding, and significantly enhances the model's ability to capture complex lip dynamics driven by speech signals. To further improve region-specific motion modeling, we introduce an audio-guided saliency splatting mechanism based on region-aware modulation, which adaptively emphasizes highly dynamic regions such as the mouth area. This allows the model to focus its learning capacity on where it matters most for accurate speech-driven animation. Extensive experiments on both the self-driven and the cross-driven settings demonstrate that our method achieves state-of-the-art visual quality, precise audio-lip synchronization, and real-time performance, outperforming prior approaches across both 2D- and 3D-based paradigms.

Jia Wan, Wanhua Li, Jason Ken Adhinarta et al.

While imaging techniques at macro and mesoscales have garnered substantial attention and resources, microscale Volume Electron Microscopy (vEM) imaging, capable of revealing intricate vascular details, has lacked the necessary benchmarking infrastructure. In this paper, we address a significant gap in this field of neuroimaging by introducing the first-in-class public benchmark, BvEM, designed specifically for cortical blood vessel segmentation in vEM images. Our BvEM benchmark is based on vEM image volumes from three mammals: adult mouse, macaque, and human. We standardized the resolution, addressed imaging variations, and meticulously annotated blood vessels through semi-automatic, manual, and quality control processes, ensuring high-quality 3D segmentation. Furthermore, we developed a zero-shot cortical blood vessel segmentation method named TriSAM, which leverages the powerful segmentation model SAM for 3D segmentation. To extend SAM from 2D to 3D volume segmentation, TriSAM employs a multi-seed tracking framework, leveraging the reliability of certain image planes for tracking while using others to identify potential turning points. This approach effectively achieves long-term 3D blood vessel segmentation without model training or fine-tuning. Experimental results show that TriSAM achieved superior performances on the BvEM benchmark across three species. Our dataset, code, and model are available online at \url{https://jia-wan.github.io/bvem}.

Pengyu Chen, Wanhua Li

In this paper, we introduce a data-efficient instance segmentation method we used in the 2021 VIPriors Instance Segmentation Challenge. Our solution is a modified version of Swin Transformer, based on the mmdetection which is a powerful toolbox. To solve the problem of lack of data, we utilize data augmentation including random flip and multiscale training to train our model. During inference, multiscale fusion is used to boost the performance. We only use a single GPU during the whole training and testing stages. In the end, our team achieved the result of 0.366 for AP@0.50:0.95 on the test set, which is competitive with other top-ranking methods while only one GPU is used. Besides, our method achieved the AP@0.50:0.95 (medium) of 0.592, which ranks second among all contestants. In the end, our team ranked third among all the contestants, as announced by the organizers.

Wanhua Li, Jiwen Lu, Abudukelimu Wuerkaixi et al.

In this paper, we investigate the problem of facial kinship verification by learning hierarchical reasoning graph networks. Conventional methods usually focus on learning discriminative features for each facial image of a paired sample and neglect how to fuse the obtained two facial image features and reason about the relations between them. To address this, we propose a Star-shaped Reasoning Graph Network (S-RGN). Our S-RGN first constructs a star-shaped graph where each surrounding node encodes the information of comparisons in a feature dimension and the central node is employed as the bridge for the interaction of surrounding nodes. Then we perform relational reasoning on this star graph with iterative message passing. The proposed S-RGN uses only one central node to analyze and process information from all surrounding nodes, which limits its reasoning capacity. We further develop a Hierarchical Reasoning Graph Network (H-RGN) to exploit more powerful and flexible capacity. More specifically, our H-RGN introduces a set of latent reasoning nodes and constructs a hierarchical graph with them. Then bottom-up comparative information abstraction and top-down comprehensive signal propagation are iteratively performed on the hierarchical graph to update the node features. Extensive experimental results on four widely used kinship databases show that the proposed methods achieve very competitive results.

Wanhua Li, Shiwei Wang, Jiwen Lu et al.

Kinship verification aims to find out whether there is a kin relation for a given pair of facial images. Kinship verification databases are born with unbalanced data. For a database with N positive kinship pairs, we naturally obtain N(N-1) negative pairs. How to fully utilize the limited positive pairs and mine discriminative information from sufficient negative samples for kinship verification remains an open issue. To address this problem, we propose a Discriminative Sample Meta-Mining (DSMM) approach in this paper. Unlike existing methods that usually construct a balanced dataset with fixed negative pairs, we propose to utilize all possible pairs and automatically learn discriminative information from data. Specifically, we sample an unbalanced train batch and a balanced meta-train batch for each iteration. Then we learn a meta-miner with the meta-gradient on the balanced meta-train batch. In the end, the samples in the unbalanced train batch are re-weighted by the learned meta-miner to optimize the kinship models. Experimental results on the widely used KinFaceW-I, KinFaceW-II, TSKinFace, and Cornell Kinship datasets demonstrate the effectiveness of the proposed approach.

Wanhua Li, Yueqi Duan, Jiwen Lu et al.

Human beings are fundamentally sociable -- that we generally organize our social lives in terms of relations with other people. Understanding social relations from an image has great potential for intelligent systems such as social chatbots and personal assistants. In this paper, we propose a simpler, faster, and more accurate method named graph relational reasoning network (GR2N) for social relation recognition. Different from existing methods which process all social relations on an image independently, our method considers the paradigm of jointly inferring the relations by constructing a social relation graph. Furthermore, the proposed GR2N constructs several virtual relation graphs to explicitly grasp the strong logical constraints among different types of social relations. Experimental results illustrate that our method generates a reasonable and consistent social relation graph and improves the performance in both accuracy and efficiency.

Wanhua Li, Yingqiang Zhang, Kangchen Lv et al.

In this paper, we propose a graph-based kinship reasoning (GKR) network for kinship verification, which aims to effectively perform relational reasoning on the extracted features of an image pair. Unlike most existing methods which mainly focus on how to learn discriminative features, our method considers how to compare and fuse the extracted feature pair to reason about the kin relations. The proposed GKR constructs a star graph called kinship relational graph where each peripheral node represents the information comparison in one feature dimension and the central node is used as a bridge for information communication among peripheral nodes. Then the GKR performs relational reasoning on this graph with recursive message passing. Extensive experimental results on the KinFaceW-I and KinFaceW-II datasets show that the proposed GKR outperforms the state-of-the-art methods.

Wanhua Li, Jiwen Lu, Jianjiang Feng et al.

Age estimation is an important yet very challenging problem in computer vision. Existing methods for age estimation usually apply a divide-and-conquer strategy to deal with heterogeneous data caused by the non-stationary aging process. However, the facial aging process is also a continuous process, and the continuity relationship between different components has not been effectively exploited. In this paper, we propose BridgeNet for age estimation, which aims to mine the continuous relation between age labels effectively. The proposed BridgeNet consists of local regressors and gating networks. Local regressors partition the data space into multiple overlapping subspaces to tackle heterogeneous data and gating networks learn continuity aware weights for the results of local regressors by employing the proposed bridge-tree structure, which introduces bridge connections into tree models to enforce the similarity between neighbor nodes. Moreover, these two components of BridgeNet can be jointly learned in an end-to-end way. We show experimental results on the MORPH II, FG-NET and Chalearn LAP 2015 datasets and find that BridgeNet outperforms the state-of-the-art methods.