Yandong Wen

Hongwei Yi, Hualin Liang, Yifei Liu et al. · amazon-science

This work addresses the problem of generating 3D holistic body motions from human speech. Given a speech recording, we synthesize sequences of 3D body poses, hand gestures, and facial expressions that are realistic and diverse. To achieve this, we first build a high-quality dataset of 3D holistic body meshes with synchronous speech. We then define a novel speech-to-motion generation framework in which the face, body, and hands are modeled separately. The separated modeling stems from the fact that face articulation strongly correlates with human speech, while body poses and hand gestures are less correlated. Specifically, we employ an autoencoder for face motions, and a compositional vector-quantized variational autoencoder (VQ-VAE) for the body and hand motions. The compositional VQ-VAE is key to generating diverse results. Additionally, we propose a cross-conditional autoregressive model that generates body poses and hand gestures, leading to coherent and realistic motions. Extensive experiments and user studies demonstrate that our proposed approach achieves state-of-the-art performance both qualitatively and quantitatively. Our novel dataset and code will be released for research purposes at https://talkshow.is.tue.mpg.de.

Radek Daněček, Kiran Chhatre, Shashank Tripathi et al. · amazon-science

To be widely adopted, 3D facial avatars must be animated easily, realistically, and directly from speech signals. While the best recent methods generate 3D animations that are synchronized with the input audio, they largely ignore the impact of emotions on facial expressions. Realistic facial animation requires lip-sync together with the natural expression of emotion. To that end, we propose EMOTE (Expressive Model Optimized for Talking with Emotion), which generates 3D talking-head avatars that maintain lip-sync from speech while enabling explicit control over the expression of emotion. To achieve this, we supervise EMOTE with decoupled losses for speech (i.e., lip-sync) and emotion. These losses are based on two key observations: (1) deformations of the face due to speech are spatially localized around the mouth and have high temporal frequency, whereas (2) facial expressions may deform the whole face and occur over longer intervals. Thus, we train EMOTE with a per-frame lip-reading loss to preserve the speech-dependent content, while supervising emotion at the sequence level. Furthermore, we employ a content-emotion exchange mechanism in order to supervise different emotions on the same audio, while maintaining the lip motion synchronized with the speech. To employ deep perceptual losses without getting undesirable artifacts, we devise a motion prior in the form of a temporal VAE. Due to the absence of high-quality aligned emotional 3D face datasets with speech, EMOTE is trained with 3D pseudo-ground-truth extracted from an emotional video dataset (i.e., MEAD). Extensive qualitative and perceptual evaluations demonstrate that EMOTE produces speech-driven facial animations with better lip-sync than state-of-the-art methods trained on the same data, while offering additional, high-quality emotional control.

Xiang Li, Yandong Wen, Muqiao Yang et al. · cmu

Previous works on voice-face matching and voice-guided face synthesis demonstrate strong correlations between voice and face, but mainly rely on coarse semantic cues such as gender, age, and emotion. In this paper, we aim to investigate the capability of reconstructing the 3D facial shape from voice from a geometry perspective without any semantic information. We propose a voice-anthropometric measurement (AM)-face paradigm, which identifies predictable facial AMs from the voice and uses them to guide 3D face reconstruction. By leveraging AMs as a proxy to link the voice and face geometry, we can eliminate the influence of unpredictable AMs and make the face geometry tractable. Our approach is evaluated on our proposed dataset with ground-truth 3D face scans and corresponding voice recordings, and we find significant correlations between voice and specific parts of the face geometry, such as the nasal cavity and cranium. Our work offers a new perspective on voice-face correlation and can serve as a good empirical study for anthropometry science.

Hao Zhang, Yao Feng, Peter Kulits et al.

Our goal is to create a realistic 3D facial avatar with hair and accessories using only a text description. While this challenge has attracted significant recent interest, existing methods either lack realism, produce unrealistic shapes, or do not support editing, such as modifications to the hairstyle. We argue that existing methods are limited because they employ a monolithic modeling approach, using a single representation for the head, face, hair, and accessories. Our observation is that the hair and face, for example, have very different structural qualities that benefit from different representations. Building on this insight, we generate avatars with a compositional model, in which the head, face, and upper body are represented with traditional 3D meshes, and the hair, clothing, and accessories with neural radiance fields (NeRF). The model-based mesh representation provides a strong geometric prior for the face region, improving realism while enabling editing of the person's appearance. By using NeRFs to represent the remaining components, our method is able to model and synthesize parts with complex geometry and appearance, such as curly hair and fluffy scarves. Our novel system synthesizes these high-quality compositional avatars from text descriptions. The experimental results demonstrate that our method, Text-guided generation and Editing of Compositional Avatars (TECA), produces avatars that are more realistic than those of recent methods while being editable because of their compositional nature. For example, our TECA enables the seamless transfer of compositional features like hairstyles, scarves, and other accessories between avatars. This capability supports applications such as virtual try-on.

Zhou Jiang, Yandong Wen, Zhen Liu

One-step text-to-image generators are attractive for deployment because they generate an image with a single forward pass, but preference finetuning them remains difficult: standard alignment methods often rely on policy likelihoods, denoising trajectories, differentiable reward gradients, or test-time optimization. We propose Drifting Preference Optimization (DrPO), an online preference-finetuning method for deterministic one-step generators. For each prompt, DrPO samples candidates from the current generator, ranks them with a target reward, and uses high- and low-scoring samples to synthesize a feature-space update direction. The update is a non-parametric dipole preference field plus a reference drift estimated from the frozen base generator, and is optimized through a detached feature-space regression target. The target reward is used only for ranking, so DrPO can train with large, black-box, or non-differentiable rewards while inference remains a single generator call. We evaluate DrPO on SD-Turbo and SDXL-Turbo with multiple target rewards and benchmarks, including HPSv3 and GenEval. DrPO improves alignment over reward-gradient-free one-step preference baselines and reduces HPSv3 training computation by $3.51\times$ under the matched effective-batch setting by removing reward-model backpropagation. Initial offline experiments suggest that sample-based gradient synthesis can also be used beyond online reward ranking.

Liao Qu, Xianwei Zou, Xiang Li et al.

This work unveils the enigmatic link between phonemes and facial features. Traditional studies on voice-face correlations typically involve using a long period of voice input, including generating face images from voices and reconstructing 3D face meshes from voices. However, in situations like voice-based crimes, the available voice evidence may be short and limited. Additionally, from a physiological perspective, each segment of speech -- phoneme -- corresponds to different types of airflow and movements in the face. Therefore, it is advantageous to discover the hidden link between phonemes and face attributes. In this paper, we propose an analysis pipeline to help us explore the voice-face relationship in a fine-grained manner, i.e., phonemes v.s. facial anthropometric measurements (AM). We build an estimator for each phoneme-AM pair and evaluate the correlation through hypothesis testing. Our results indicate that AMs are more predictable from vowels compared to consonants, particularly with plosives. Additionally, we observe that if a specific AM exhibits more movement during phoneme pronunciation, it is more predictable. Our findings support those in physiology regarding correlation and lay the groundwork for future research on speech-face multimodal learning.

Weiyang Liu, Zeju Qiu, Yao Feng et al.

Large foundation models are becoming ubiquitous, but training them from scratch is prohibitively expensive. Thus, efficiently adapting these powerful models to downstream tasks is increasingly important. In this paper, we study a principled finetuning paradigm -- Orthogonal Finetuning (OFT) -- for downstream task adaptation. Despite demonstrating good generalizability, OFT still uses a fairly large number of trainable parameters due to the high dimensionality of orthogonal matrices. To address this, we start by examining OFT from an information transmission perspective, and then identify a few key desiderata that enable better parameter-efficiency. Inspired by how the Cooley-Tukey fast Fourier transform algorithm enables efficient information transmission, we propose an efficient orthogonal parameterization using butterfly structures. We apply this parameterization to OFT, creating a novel parameter-efficient finetuning method, called Orthogonal Butterfly (BOFT). By subsuming OFT as a special case, BOFT introduces a generalized orthogonal finetuning framework. Finally, we conduct an extensive empirical study of adapting large vision transformers, large language models, and text-to-image diffusion models to various downstream tasks in vision and language.

Yangyi Huang, Ruotian Peng, Zeju Qiu et al.

Parameter-efficient finetuning (PEFT) has become the standard approach for adapting large language models, yet evaluations largely emphasize downstream accuracy while overlooking the retention of pretrained capabilities. We argue that PEFT should be assessed through the stability-plasticity dilemma: the trade-off between target-task adaptation and resistance to forgetting. We introduce PEFT-Arena, a benchmark that jointly measures downstream performance and general capability retention. Across methods, we find distinct stability-plasticity profiles; under comparable parameter budgets, orthogonal finetuning achieves the most favorable Pareto frontier. To explain these differences, we analyze PEFT updates from two geometric perspectives. In weight space, spectral analysis reveals how parameterizations interact with the pretrained singular-value structure. In activation space, retention metrics show whether finetuning preserves or distorts general-capability representations, with forgetting linked to non-isometric representation distortion. Finally, an analysis shows that final SFT checkpoints often overshoot a better target-retention operating point. Inspired by this, we present case studies of a post-hoc improvement with path-wise rewinding.

Yandong Wen, Weiyang Liu, Yao Feng et al.

In this paper, we focus on a general yet important learning problem, pairwise similarity learning (PSL). PSL subsumes a wide range of important applications, such as open-set face recognition, speaker verification, image retrieval and person re-identification. The goal of PSL is to learn a pairwise similarity function assigning a higher similarity score to positive pairs (i.e., a pair of samples with the same label) than to negative pairs (i.e., a pair of samples with different label). We start by identifying a key desideratum for PSL, and then discuss how existing methods can achieve this desideratum. We then propose a surprisingly simple proxy-free method, called SimPLE, which requires neither feature/proxy normalization nor angular margin and yet is able to generalize well in open-set recognition. We apply the proposed method to three challenging PSL tasks: open-set face recognition, image retrieval and speaker verification. Comprehensive experimental results on large-scale benchmarks show that our method performs significantly better than current state-of-the-art methods.

Jinxiu Liu, Xuanming Liu, Kangfu Mei et al.

Large pretrained diffusion models have significantly enhanced the quality of generated videos, and yet their use in real-time streaming remains limited. Autoregressive models offer a natural framework for sequential frame synthesis but require heavy computation to achieve high fidelity. Diffusion distillation can compress these models into efficient few-step variants, but existing video distillation approaches largely adapt image-specific methods that neglect temporal dependencies. These techniques often excel in image generation but underperform in video synthesis, exhibiting reduced motion coherence, error accumulation over long sequences, and a latency-quality trade-off. We identify two factors that result in these limitations: insufficient utilization of temporal context during step reduction and implicit prediction of subsequent noise levels in next-chunk prediction (i.e., exposure bias). To address these issues, we propose Diagonal Distillation, which operates orthogonally to existing approaches and better exploits temporal information across both video chunks and denoising steps. Central to our approach is an asymmetric generation strategy: more steps early, fewer steps later. This design allows later chunks to inherit rich appearance information from thoroughly processed early chunks, while using partially denoised chunks as conditional inputs for subsequent synthesis. By aligning the implicit prediction of subsequent noise levels during chunk generation with the actual inference conditions, our approach mitigates error propagation and reduces oversaturation in long-range sequences. We further incorporate implicit optical flow modeling to preserve motion quality under strict step constraints. Our method generates a 5-second video in 2.61 seconds (up to 31 FPS), achieving a 277.3x speedup over the undistilled model.

Yifei Liu, Qiong Cao, Yandong Wen et al.

This paper addresses the problem of generating lifelike holistic co-speech motions for 3D avatars, focusing on two key aspects: variability and coordination. Variability allows the avatar to exhibit a wide range of motions even with similar speech content, while coordination ensures a harmonious alignment among facial expressions, hand gestures, and body poses. We aim to achieve both with ProbTalk, a unified probabilistic framework designed to jointly model facial, hand, and body movements in speech. ProbTalk builds on the variational autoencoder (VAE) architecture and incorporates three core designs. First, we introduce product quantization (PQ) to the VAE, which enriches the representation of complex holistic motion. Second, we devise a novel non-autoregressive model that embeds 2D positional encoding into the product-quantized representation, thereby preserving essential structure information of the PQ codes. Last, we employ a secondary stage to refine the preliminary prediction, further sharpening the high-frequency details. Coupling these three designs enables ProbTalk to generate natural and diverse holistic co-speech motions, outperforming several state-of-the-art methods in qualitative and quantitative evaluations, particularly in terms of realism. Our code and model will be released for research purposes at https://feifeifeiliu.github.io/probtalk/.

Ruotian Peng, Haiying He, Yake Wei et al.

High-quality image captions play a crucial role in improving the performance of cross-modal applications such as text-to-image generation, text-to-video generation, and text-image retrieval. To generate long-form, high-quality captions, many recent studies have employed multimodal large language models (MLLMs). However, current MLLMs often produce captions that lack fine-grained details or suffer from hallucinations, a challenge that persists in both open-source and closed-source models. Inspired by Feature-Integration theory, which suggests that attention must focus on specific regions to integrate visual information effectively, we propose a \textbf{divide-then-aggregate} strategy. Our method first divides the image into semantic and spatial patches to extract fine-grained details, enhancing the model's local perception of the image. These local details are then hierarchically aggregated to generate a comprehensive global description. To address hallucinations and inconsistencies in the generated captions, we apply a semantic-level filtering process during hierarchical aggregation. This training-free pipeline can be applied to both open-source models (LLaVA-1.5, LLaVA-1.6, Mini-Gemini) and closed-source models (Claude-3.5-Sonnet, GPT-4o, GLM-4V-Plus). Extensive experiments demonstrate that our method generates more detailed, reliable captions, advancing multimodal description generation without requiring model retraining. The source code are available at https://github.com/GeWu-Lab/Patch-Matters

Kexuan Shi, Hanxuan Li, Zeju Qiu et al.

We introduce Pion, a spectrum-preserving optimizer for large language model (LLM) training based on orthogonal equivalence transformation. Unlike additive optimizers such as Adam and Muon, Pion updates each weight matrix through left and right orthogonal transformations, preserving its singular values throughout training. This yields an optimization mechanism that modulates the geometry of weight matrices while keeping their spectral norm fixed. We derive the Pion update rule, systematically examine its design choices, and analyze its convergence behavior along with several key properties. Empirical results show that Pion offers a stable and competitive alternative to standard optimizers for both LLM pretraining and finetuning.

Yamei Chen, Haoquan Zhang, Yangyi Huang et al.

Large language models (LLMs) excel at program synthesis, yet their ability to produce symbolic graphics programs (SGPs) that render into precise visual content remains underexplored. We study symbolic graphics programming, where the goal is to generate an SGP from a natural-language description. This task also serves as a lens into how LLMs understand the visual world by prompting them to generate images rendered from SGPs. Among various SGPs, our paper sticks to scalable vector graphics (SVGs). We begin by examining the extent to which LLMs can generate SGPs. To this end, we introduce SGP-GenBench, a comprehensive benchmark covering object fidelity, scene fidelity, and compositionality (attribute binding, spatial relations, numeracy). On SGP-GenBench, we discover that frontier proprietary models substantially outperform open-source models, and performance correlates well with general coding capabilities. Motivated by this gap, we aim to improve LLMs' ability to generate SGPs. We propose a reinforcement learning (RL) with verifiable rewards approach, where a format-validity gate ensures renderable SVG, and a cross-modal reward aligns text and the rendered image via strong vision encoders (e.g., SigLIP for text-image and DINO for image-image). Applied to Qwen-2.5-7B, our method substantially improves SVG generation quality and semantics, achieving performance on par with frontier systems. We further analyze training dynamics, showing that RL induces (i) finer decomposition of objects into controllable primitives and (ii) contextual details that improve scene coherence. Our results demonstrate that symbolic graphics programming offers a precise and interpretable lens on cross-modal grounding.

Alexander Richard, Michael Zollhoefer, Yandong Wen et al.

This paper presents a generic method for generating full facial 3D animation from speech. Existing approaches to audio-driven facial animation exhibit uncanny or static upper face animation, fail to produce accurate and plausible co-articulation or rely on person-specific models that limit their scalability. To improve upon existing models, we propose a generic audio-driven facial animation approach that achieves highly realistic motion synthesis results for the entire face. At the core of our approach is a categorical latent space for facial animation that disentangles audio-correlated and audio-uncorrelated information based on a novel cross-modality loss. Our approach ensures highly accurate lip motion, while also synthesizing plausible animation of the parts of the face that are uncorrelated to the audio signal, such as eye blinks and eye brow motion. We demonstrate that our approach outperforms several baselines and obtains state-of-the-art quality both qualitatively and quantitatively. A perceptual user study demonstrates that our approach is deemed more realistic than the current state-of-the-art in over 75% of cases. We recommend watching the supplemental video before reading the paper: https://github.com/facebookresearch/meshtalk

Zhouliang Yu, Ruotian Peng, Keyi Ding et al.

Formal mathematical reasoning remains a critical challenge for artificial intelligence, hindered by limitations of existing benchmarks in scope and scale. To address this, we present FormalMATH, a large-scale Lean4 benchmark comprising 5,560 formally verified problems spanning from high-school Olympiad challenges to undergraduate-level theorems across diverse domains (e.g., algebra, applied mathematics, calculus, number theory, and discrete mathematics). To mitigate the inefficiency of manual formalization, we introduce a novel human-in-the-loop autoformalization pipeline that integrates: (1) specialized large language models (LLMs) for statement autoformalization, (2) multi-LLM semantic verification, and (3) negation-based disproof filtering strategies using off-the-shelf LLM-based provers. This approach reduces expert annotation costs by retaining 72.09% of statements before manual verification while ensuring fidelity to the original natural-language problems. Our evaluation of state-of-the-art LLM-based theorem provers reveals significant limitations: even the strongest models achieve only 16.46% success rate under practical sampling budgets, exhibiting pronounced domain bias (e.g., excelling in algebra but failing in calculus) and over-reliance on simplified automation tactics. Notably, we identify a counterintuitive inverse relationship between natural-language solution guidance and proof success in chain-of-thought reasoning scenarios, suggesting that human-written informal reasoning introduces noise rather than clarity in the formal reasoning settings. We believe that FormalMATH provides a robust benchmark for benchmarking formal mathematical reasoning.

Ruotian Peng, Yi Ren, Zhouliang Yu et al.

Reinforcement learning with verifiable rewards (RLVR) has advanced the reasoning capabilities of large language models (LLMs). However, prevailing RLVR methods exhibit a systematic bias toward exploitation over exploration, as evidenced by improved pass@1 but reduced pass@K (K>1) performance. To understand this issue, we analyze training dynamics of RLVR methods by tracking the token-level probability distributions over vocabulary candidates. Our analysis reveals a consistent probability concentration effect where the top-1 candidate increasingly accumulates probability mass and suppresses that of other candidates. More importantly, stronger over-concentration correlates with worse pass@K performance. Inspired by this finding, we propose Simple Pass@K Optimization (SimKO), a method designed to mitigate the over-concentration issue, thereby encouraging exploration. SimKO operates in an asymmetrical manner. For verified-correct responses, it boosts the probabilities of the top-K candidates. For verified-incorrect responses, it applies stronger penalties to the top-1 candidate. We observe that this asymmetric design is particularly effective at mitigating over-concentration when applied at tokens with high entropy. Across various math and logical-reasoning benchmarks, SimKO consistently yields higher pass@K for a wide range of K, providing a simple way to improve RLVR's exploration.

Zhou Jiang, Yandong Wen, Zhen Liu

Aligning large-scale text-to-image diffusion models with nuanced human preferences remains challenging. While direct preference optimization (DPO) is simple and effective, large-scale finetuning often shows a generalization gap. We take inspiration from test-time guidance and cast preference alignment as classifier-free guidance (CFG): a finetuned preference model acts as an external control signal during sampling. Building on this view, we propose a simple method that improves alignment without retraining the base model. To further enhance generalization, we decouple preference learning into two modules trained on positive and negative data, respectively, and form a \emph{contrastive guidance} vector at inference by subtracting their predictions (positive minus negative), scaled by a user-chosen strength and added to the base prediction at each step. This yields a sharper and controllable alignment signal. We evaluate on Stable Diffusion 1.5 and Stable Diffusion XL with Pick-a-Pic v2 and HPDv3, showing consistent quantitative and qualitative gains.

Kexuan Shi, Yandong Wen, Weiyang Liu

Model merging is an efficient post-training strategy for integrating knowledge from multiple finetuned checkpoints of a shared foundation model. Existing methods operate in the parameter space, combining task vectors to mitigate conflicts, but remain constrained by parameter inconsistencies. We propose Functional Dual Anchors (FDAs), a framework that instead models the input-representation space. FDAs are synthetic inputs whose induced gradients align with task vectors, capturing task-specific functional shifts relative to the pretrained model. This perspective bridges joint multi-task training and post-hoc merging, offering both robustness and flexibility. We further introduce a principled initialization scheme and show that FDAs are complementary to parameter-space model merging. Comprehensive experiments demonstrate the effectiveness of FDAs in model merging.

Yandong Wen, Weiyang Liu, Bhiksha Raj et al.

We present a conditional estimation (CEST) framework to learn 3D facial parameters from 2D single-view images by self-supervised training from videos. CEST is based on the process of analysis by synthesis, where the 3D facial parameters (shape, reflectance, viewpoint, and illumination) are estimated from the face image, and then recombined to reconstruct the 2D face image. In order to learn semantically meaningful 3D facial parameters without explicit access to their labels, CEST couples the estimation of different 3D facial parameters by taking their statistical dependency into account. Specifically, the estimation of any 3D facial parameter is not only conditioned on the given image, but also on the facial parameters that have already been derived. Moreover, the reflectance symmetry and consistency among the video frames are adopted to improve the disentanglement of facial parameters. Together with a novel strategy for incorporating the reflectance symmetry and consistency, CEST can be efficiently trained with in-the-wild video clips. Both qualitative and quantitative experiments demonstrate the effectiveness of CEST.

Weiyang Liu, Yandong Wen, Bhiksha Raj et al.

This paper addresses the deep face recognition problem under an open-set protocol, where ideal face features are expected to have smaller maximal intra-class distance than minimal inter-class distance under a suitably chosen metric space. To this end, hyperspherical face recognition, as a promising line of research, has attracted increasing attention and gradually become a major focus in face recognition research. As one of the earliest works in hyperspherical face recognition, SphereFace explicitly proposed to learn face embeddings with large inter-class angular margin. However, SphereFace still suffers from severe training instability which limits its application in practice. In order to address this problem, we introduce a unified framework to understand large angular margin in hyperspherical face recognition. Under this framework, we extend the study of SphereFace and propose an improved variant with substantially better training stability -- SphereFace-R. Specifically, we propose two novel ways to implement the multiplicative margin, and study SphereFace-R under three different feature normalization schemes (no feature normalization, hard feature normalization and soft feature normalization). We also propose an implementation strategy -- "characteristic gradient detachment" -- to stabilize training. Extensive experiments on SphereFace-R show that it is consistently better than or competitive with state-of-the-art methods.

Yandong Wen, Weiyang Liu, Adrian Weller et al.

State-of-the-art deep face recognition methods are mostly trained with a softmax-based multi-class classification framework. Despite being popular and effective, these methods still have a few shortcomings that limit empirical performance. In this paper, we start by identifying the discrepancy between training and evaluation in the existing multi-class classification framework and then discuss the potential limitations caused by the "competitive" nature of softmax normalization. Motivated by these limitations, we propose a novel binary classification training framework, termed SphereFace2. In contrast to existing methods, SphereFace2 circumvents the softmax normalization, as well as the corresponding closed-set assumption. This effectively bridges the gap between training and evaluation, enabling the representations to be improved individually by each binary classification task. Besides designing a specific well-performing loss function, we summarize a few general principles for this "one-vs-all" binary classification framework so that it can outperform current competitive methods. Our experiments on popular benchmarks demonstrate that SphereFace2 can consistently outperform state-of-the-art deep face recognition methods. The code has been made publicly available.

Yandong Wen, Rita Singh, Bhiksha Raj

Voice profiling aims at inferring various human parameters from their speech, e.g. gender, age, etc. In this paper, we address the challenge posed by a subtask of voice profiling - reconstructing someone's face from their voice. The task is designed to answer the question: given an audio clip spoken by an unseen person, can we picture a face that has as many common elements, or associations as possible with the speaker, in terms of identity? To address this problem, we propose a simple but effective computational framework based on generative adversarial networks (GANs). The network learns to generate faces from voices by matching the identities of generated faces to those of the speakers, on a training set. We evaluate the performance of the network by leveraging a closely related task - cross-modal matching. The results show that our model is able to generate faces that match several biometric characteristics of the speaker, and results in matching accuracies that are much better than chance.

Yandong Wen, Mahmoud Al Ismail, Weiyang Liu et al.

We propose a novel framework, called Disjoint Mapping Network (DIMNet), for cross-modal biometric matching, in particular of voices and faces. Different from the existing methods, DIMNet does not explicitly learn the joint relationship between the modalities. Instead, DIMNet learns a shared representation for different modalities by mapping them individually to their common covariates. These shared representations can then be used to find the correspondences between the modalities. We show empirically that DIMNet is able to achieve better performance than other current methods, with the additional benefits of being conceptually simpler and less data-intensive.

Yandong Wen, Mahmoud Al Ismail, Bhiksha Raj et al.

In many retrieval problems, where we must retrieve one or more entries from a gallery in response to a probe, it is common practice to learn to do by directly comparing the probe and gallery entries to one another. In many situations the gallery and probe have common covariates -- external variables that are common to both. In principle it is possible to perform the retrieval based merely on these covariates. The process, however, becomes gated by our ability to recognize the covariates for the probe and gallery entries correctly. In this paper we analyze optimal strategies for retrieval based only on matching covariates, when the recognition of the covariates is itself inaccurate. We investigate multiple problems: recovering one item from a gallery of $N$ entries, matching pairs of instances, and retrieval from large collections. We verify our analytical formulae through experiments to verify their correctness in practical settings.

Weiyang Liu, Yandong Wen, Zhiding Yu et al.

This paper addresses deep face recognition (FR) problem under open-set protocol, where ideal face features are expected to have smaller maximal intra-class distance than minimal inter-class distance under a suitably chosen metric space. However, few existing algorithms can effectively achieve this criterion. To this end, we propose the angular softmax (A-Softmax) loss that enables convolutional neural networks (CNNs) to learn angularly discriminative features. Geometrically, A-Softmax loss can be viewed as imposing discriminative constraints on a hypersphere manifold, which intrinsically matches the prior that faces also lie on a manifold. Moreover, the size of angular margin can be quantitatively adjusted by a parameter $m$. We further derive specific $m$ to approximate the ideal feature criterion. Extensive analysis and experiments on Labeled Face in the Wild (LFW), Youtube Faces (YTF) and MegaFace Challenge show the superiority of A-Softmax loss in FR tasks. The code has also been made publicly available.

Weiyang Liu, Yandong Wen, Zhiding Yu et al.

Cross-entropy loss together with softmax is arguably one of the most common used supervision components in convolutional neural networks (CNNs). Despite its simplicity, popularity and excellent performance, the component does not explicitly encourage discriminative learning of features. In this paper, we propose a generalized large-margin softmax (L-Softmax) loss which explicitly encourages intra-class compactness and inter-class separability between learned features. Moreover, L-Softmax not only can adjust the desired margin but also can avoid overfitting. We also show that the L-Softmax loss can be optimized by typical stochastic gradient descent. Extensive experiments on four benchmark datasets demonstrate that the deeply-learned features with L-softmax loss become more discriminative, hence significantly boosting the performance on a variety of visual classification and verification tasks.

Xiao Zhang, Zhiyuan Fang, Yandong Wen et al.

Convolutional neural networks have achieved great improvement on face recognition in recent years because of its extraordinary ability in learning discriminative features of people with different identities. To train such a well-designed deep network, tremendous amounts of data is indispensable. Long tail distribution specifically refers to the fact that a small number of generic entities appear frequently while other objects far less existing. Considering the existence of long tail distribution of the real world data, large but uniform distributed data are usually hard to retrieve. Empirical experiences and analysis show that classes with more samples will pose greater impact on the feature learning process and inversely cripple the whole models feature extracting ability on tail part data. Contrary to most of the existing works that alleviate this problem by simply cutting the tailed data for uniform distributions across the classes, this paper proposes a new loss function called range loss to effectively utilize the whole long tailed data in training process. More specifically, range loss is designed to reduce overall intra-personal variations while enlarging inter-personal differences within one mini-batch simultaneously when facing even extremely unbalanced data. The optimization objective of range loss is the $k$ greatest range's harmonic mean values in one class and the shortest inter-class distance within one batch. Extensive experiments on two famous and challenging face recognition benchmarks (Labeled Faces in the Wild (LFW) and YouTube Faces (YTF) not only demonstrate the effectiveness of the proposed approach in overcoming the long tail effect but also show the good generalization ability of the proposed approach.

Weiyang Liu, Zhiding Yu, Yandong Wen et al.

Sparse coding with dictionary learning (DL) has shown excellent classification performance. Despite the considerable number of existing works, how to obtain features on top of which dictionaries can be better learned remains an open and interesting question. Many current prevailing DL methods directly adopt well-performing crafted features. While such strategy may empirically work well, it ignores certain intrinsic relationship between dictionaries and features. We propose a framework where features and dictionaries are jointly learned and optimized. The framework, named joint non-negative projection and dictionary learning (JNPDL), enables interaction between the input features and the dictionaries. The non-negative projection leads to discriminative parts-based object features while DL seeks a more suitable representation. Discriminative graph constraints are further imposed to simultaneously maximize intra-class compactness and inter-class separability. Experiments on both image and image set classification show the excellent performance of JNPDL by outperforming several state-of-the-art approaches.

Yandong Wen, Weiyang Liu, Meng Yang et al.

Practical face recognition has been studied in the past decades, but still remains an open challenge. Current prevailing approaches have already achieved substantial breakthroughs in recognition accuracy. However, their performance usually drops dramatically if face samples are severely misaligned. To address this problem, we propose a highly efficient misalignment-robust locality-constrained representation (MRLR) algorithm for practical real-time face recognition. Specifically, the locality constraint that activates the most correlated atoms and suppresses the uncorrelated ones, is applied to construct the dictionary for face alignment. Then we simultaneously align the warped face and update the locality-constrained dictionary, eventually obtaining the final alignment. Moreover, we make use of the block structure to accelerate the derived analytical solution. Experimental results on public data sets show that MRLR significantly outperforms several state-of-the-art approaches in terms of efficiency and scalability with even better performance.

Yandong Wen, Weiyang Liu, Meng Yang et al.

Occlusion in face recognition is a common yet challenging problem. While sparse representation based classification (SRC) has been shown promising performance in laboratory conditions (i.e. noiseless or random pixel corrupted), it performs much worse in practical scenarios. In this paper, we consider the practical face recognition problem, where the occlusions are predictable and available for sampling. We propose the structured occlusion coding (SOC) to address occlusion problems. The structured coding here lies in two folds. On one hand, we employ a structured dictionary for recognition. On the other hand, we propose to use the structured sparsity in this formulation. Specifically, SOC simultaneously separates the occlusion and classifies the image. In this way, the problem of recognizing an occluded image is turned into seeking a structured sparse solution on occlusion-appended dictionary. In order to construct a well-performing occlusion dictionary, we propose an occlusion mask estimating technique via locality constrained dictionary (LCD), showing striking improvement in occlusion sample. On a category-specific occlusion dictionary, we replace norm sparsity with the structured sparsity which is shown more robust, further enhancing the robustness of our approach. Moreover, SOC achieves significant improvement in handling large occlusion in real world. Extensive experiments are conducted on public data sets to validate the superiority of the proposed algorithm.

Weiyang Liu, Zhiding Yu, Lijia Lu et al.

We consider the image classification problem via kernel collaborative representation classification with locality constrained dictionary (KCRC-LCD). Specifically, we propose a kernel collaborative representation classification (KCRC) approach in which kernel method is used to improve the discrimination ability of collaborative representation classification (CRC). We then measure the similarities between the query and atoms in the global dictionary in order to construct a locality constrained dictionary (LCD) for KCRC. In addition, we discuss several similarity measure approaches in LCD and further present a simple yet effective unified similarity measure whose superiority is validated in experiments. There are several appealing aspects associated with LCD. First, LCD can be nicely incorporated under the framework of KCRC. The LCD similarity measure can be kernelized under KCRC, which theoretically links CRC and LCD under the kernel method. Second, KCRC-LCD becomes more scalable to both the training set size and the feature dimension. Example shows that KCRC is able to perfectly classify data with certain distribution, while conventional CRC fails completely. Comprehensive experiments on many public datasets also show that KCRC-LCD is a robust discriminative classifier with both excellent performance and good scalability, being comparable or outperforming many other state-of-the-art approaches.