Universal multimodal embedding models are essential in various tasks. Existing approaches typically use in-batch mining to identify hard negatives by measuring the similarity of query-candidate pairs. However, these methods often struggle to capture subtle semantic differences among candidates and lack diversity in negative samples. Moreover, the embeddings exhibit limited discriminative ability in distinguishing false and hard negatives. In this paper, we leverage the advanced understanding capabilities of MLLMs to enhance representation learning, and present a novel Universal Multimodal Embedding(UniME-V2) model. Our approach first constructs a potential hard negative set through global retrieval. We then introduce the MLLM-as-a-Judge mechanism, which utilizes MLLMs to assess the semantic alignment of query-candidate pairs and generate soft semantic matching scores. These scores serve as a foundation for hard negative mining, mitigating the impact of false negatives and enabling the identification of diverse, high-quality hard negatives. Furthermore, the semantic matching scores are used as soft labels to mitigate the rigid one-to-one mapping constraint. By aligning the similarity matrix with the soft semantic matching score matrix, the model learns semantic distinctions among candidates, significantly enhancing its discriminative capacity. To further improve performance, we propose UniME-V2, a reranking model trained on our mined hard negatives through a joint pairwise and listwise optimization approach. We conduct comprehensive experiments on the MMEB benchmark and multiple retrieval tasks, demonstrating that our method achieves state-of-the-art performance across all tasks.

Large Multimodal Models (LMMs) often face a modality representation gap during pretraining: while language embeddings remain stable, visual representations are highly sensitive to contextual noise (e.g., background clutter). To address this issue, we introduce a visual comprehension stage, which we call ViCToR (Visual Comprehension via Token Reconstruction), a novel pretraining framework for LMMs. ViCToR employs a learnable visual token pool and utilizes the Hungarian matching algorithm to select semantically relevant tokens from this pool for visual token replacement. Furthermore, by integrating a visual token reconstruction loss with dense semantic supervision, ViCToR can learn tokens which retain high visual detail, thereby enhancing the large language model's (LLM's) understanding of visual information. After pretraining on 3 million publicly accessible images and captions, ViCToR achieves state-of-the-art results, improving over LLaVA-NeXT-8B by 10.4%, 3.2%, and 7.2% on the MMStar, SEEDI, and RealWorldQA benchmarks, respectively.

The performance of egocentric AI agents is fundamentally limited by multimodal intent ambiguity. This challenge arises from a combination of underspecified language, imperfect visual data, and deictic gestures, which frequently leads to task failure. Existing monolithic Vision-Language Models (VLMs) struggle to resolve these multimodal ambiguous inputs, often failing silently or hallucinating responses. To address these ambiguities, we introduce the Plug-and-Play Clarifier, a zero-shot and modular framework that decomposes the problem into discrete, solvable sub-tasks. Specifically, our framework consists of three synergistic modules: (1) a text clarifier that uses dialogue-driven reasoning to interactively disambiguate linguistic intent, (2) a vision clarifier that delivers real-time guidance feedback, instructing users to adjust their positioning for improved capture quality, and (3) a cross-modal clarifier with grounding mechanism that robustly interprets 3D pointing gestures and identifies the specific objects users are pointing to. Extensive experiments demonstrate that our framework improves the intent clarification performance of small language models (4-8B) by approximately 30%, making them competitive with significantly larger counterparts. We also observe consistent gains when applying our framework to these larger models. Furthermore, our vision clarifier increases corrective guidance accuracy by over 20%, and our cross-modal clarifier improves semantic answer accuracy for referential grounding by 5%. Overall, our method provides a plug-and-play framework that effectively resolves multimodal ambiguity and significantly enhances user experience in egocentric interaction.

What if accessing the web did not require a screen, a stable desk, or even free hands? For people navigating crowded cities, living with low vision, or experiencing cognitive overload, smart glasses coupled with AI agents could turn the web into an always-on assistive layer over daily life. We present Egocentric Co-Pilot, a web-native neuro-symbolic framework that runs on smart glasses and uses a Large Language Model (LLM) to orchestrate a toolbox of perception, reasoning, and web tools. An egocentric reasoning core combines Temporal Chain-of-Thought with Hierarchical Context Compression to support long-horizon question answering and decision support over continuous first-person video, far beyond a single model's context window. Additionally, a lightweight multimodal intent layer maps noisy speech and gaze into structured commands. We further implement and evaluate a cloud-native WebRTC pipeline integrating streaming speech, video, and control messages into a unified channel for smart glasses and browsers. In parallel, we deploy an on-premise WebSocket baseline, exposing concrete trade-offs between local inference and cloud offloading in terms of latency, mobility, and resource use. Experiments on Egolife and HD-EPIC demonstrate competitive or state-of-the-art egocentric QA performance, and a human-in-the-loop study on smart glasses shows higher task completion and user satisfaction than leading commercial baselines. Taken together, these results indicate that web-connected egocentric co-pilots can be a practical path toward more accessible, context-aware assistance in everyday life. By grounding operation in web-native communication primitives and modular, auditable tool use, Egocentric Co-Pilot offers a concrete blueprint for assistive, always-on web agents that support education, accessibility, and social inclusion for people who may benefit most from contextual, egocentric AI.

Abstract Existing Graph Neural Networks (GNNs) are limited to process graphs each of whose vertices is represented by a vector or a single value, limited their representing capability to describe complex objects. In this paper, we propose a novel GNN (called Graph in Graph Neural (GIG) Network) which can process graph-style data (called GIG sample) whose vertices are further represented by graphs. Given a set of graphs or a data sample whose components can be represented by a set of graphs (called multi-graph data sample), our GIG network starts with a GIG sample generation (GSG) module which encodes the input as a GIG sample , where each GIG vertex includes a graph. Then, a set of GIG hidden layers are stacked, with each consisting of: (1) a GIG vertex-level updating (GVU) module that individually updates the graph in every GIG vertex based on its internal information; and (2) a global-level GIG sample updating (GGU) module that updates graphs in all GIG vertices based on their relationships, making the updated GIG vertices become global context-aware. This way, both internal cues within the graph contained in each GIG vertex and the relationships among GIG vertices could be utilized for down-stream tasks. Experimental results demonstrate that our GIG network generalizes well for not only various generic graph analysis tasks but also real-world multi-graph data analysis (e.g., human skeleton video-based action recognition), which achieved the new state-of-the-art results on 15 out of 16 evaluated datasets. Our code is publicly available at https://github.com/wangjs96/Graph-in-Graph-Neural-Network .

Face Recognition (FR) technology has made significant strides with the emergence of deep learning. Typically, most existing FR models are built upon Convolutional Neural Networks (CNN) and take RGB face images as the model's input. In this work, we take a closer look at existing FR paradigms from high-efficiency, security, and precision perspectives, and identify the following three problems: (i) CNN frameworks are vulnerable in capturing global facial features and modeling the correlations between local facial features. (ii) Selecting RGB face images as the model's input greatly degrades the model's inference efficiency, increasing the extra computation costs. (iii) In the real-world FR system that operates on RGB face images, the integrity of user privacy may be compromised if hackers successfully penetrate and gain access to the input of this model. To solve these three issues, we propose two novel FR frameworks, i.e., TransFace and TransFace++, which successfully explore the feasibility of applying ViTs and image bytes to FR tasks, respectively. Firstly, as revealed from our observations, we find that ViTs perform vulnerably when applied to FR scenarios with extremely large datasets. We investigate the reasons for this phenomenon and discover that the existing data augmentation approaches and hard sample mining strategies are incompatible with ViTs-based FR backbone due to the lack of tailored consideration on preserving face structural information and leveraging each local token information. To remedy these problems, we first propose a superior FR model called TransFace, which contains a patch-level data augmentation strategy named Dominant Patch Amplitude Perturbation (DPAP) and a hard sample mining strategy named Entropy-guided Hard Sample Mining (EHSM). Furthermore, to improve inference efficiency and user privacy protection, we investigate the intrinsic property of image bytes and propose a superior FR model termed TransFace++. The proposed model is trained directly on image bytes, presenting a novel approach to address the aforementioned issues. Specifically, considering the importance of local correlations in bytes, an image bytes compression strategy named Topology-based Image Bytes Compression (TIBC) is introduced to extract prominent features from the raw bytes and integrate these features with byte embeddings, effectively mitigating information loss during the bytes mapping process. Moreover, to strengthen the model's perception on geometric information encoded in image bytes, a novel cross-attention module named Structure Information-guided Cross-Attention (SICA) is designed to inject structure information into byte tokens for information interaction, significantly improving the model's generalization ability. Experiments on popular face benchmarks demonstrate the superiority of our TransFace and TransFace++.

Implicit surface representations such as the signed distance function (SDF) have emerged as a promising approach for image-based surface reconstruction. However, existing optimization methods assume opaque surfaces and therefore cannot properly reconstruct translucent surfaces and sub-pixel thin structures, which also exhibit low opacity due to the blending effect. While neural radiance field (NeRF) based methods can model semi-transparency and synthesize novel views with photo-realistic quality, their volumetric representation tightly couples geometry (surface occupancy) and material property (surface opacity), and therefore cannot be easily converted into surfaces without introducing artifacts. We present α Surf, a novel scene representation with decoupled geometry and opacity for the reconstruction of surfaces with translucent or blending effects. Ray-surface intersections on our representation can be found in closed-form via analytical solutions of cubic polynomials, avoiding Monte-Carlo sampling, and are fully differentiable by construction. Our qualitative and quantitative evaluations show that our approach can accurately reconstruct translucent and extremely thin surfaces, achieving better reconstruction quality than state-of-the-art SDF and NeRF methods.

Contrastive Language-Image Pre-training (CLIP) has achieved excellent performance over a wide range of tasks. However, the effectiveness of CLIP heavily relies on a substantial corpus of pre-training data, resulting in notable consumption of computational resources. Although knowledge distillation has been widely applied in single modality models, how to efficiently expand knowledge distillation to vision-language foundation models with extensive data remains relatively unexplored. In this paper, we introduce CLIP-CID, a novel distillation mechanism that effectively transfers knowledge from a large vision-language foundation model to a smaller model. We initially propose a simple but efficient image semantic balance method to reduce transfer learning bias and improve distillation efficiency. This method filters out 43.7% of image-text pairs from the LAION400M while maintaining superior performance. After that, we leverage cluster-instance discrimination to facilitate knowledge transfer from the teacher model to the student model, thereby empowering the student model to acquire a holistic semantic comprehension of the pre-training data. Experimental results demonstrate that CLIP-CID achieves state-of-the-art performance on various downstream tasks including linear probe and zero-shot classification.

Audio-Visual Segmentation (AVS) aims to identify, at the pixel level, the object in a visual scene that produces a given sound. Current AVS methods rely on costly fine-grained annotations of mask-audio pairs, making them impractical for scalability. To address this, we propose the Modality Correspondence Alignment (MoCA) framework, which seamlessly integrates off-the-shelf foundation models like DINO, SAM, and ImageBind. Our approach leverages existing knowledge within these models and optimizes their joint usage for multimodal associations. Our approach relies on estimating positive and negative image pairs in the feature space. For pixel-level association, we introduce an audio-visual adapter and a novel {pixel matching aggregation} strategy within the image-level contrastive learning framework. This allows for a flexible connection between object appearance and audio signal at the pixel level, with tolerance to imaging variations such as translation and rotation. Extensive experiments on the AVSBench (single and multi-object splits) and AVSS datasets demonstrate that MoCA outperforms unsupervised baseline approaches and some supervised counterparts, particularly in complex scenarios with multiple auditory objects. In terms of mIoU, MoCA achieves a substantial improvement over baselines in both the AVSBench (S4: +17.24%, MS3: +67.64%) and AVSS (+19.23%) audio-visual segmentation challenges.

Hand-object interaction modeling from a single RGB image is a significantly challenging task. Previous works typically reconstruct hand-object interactions as texture-less meshes, ignoring photo-realistic image generation. In this work, we introduce the HO123, a novel method to synthesize novel-view hand-object interaction images from a single image. To this end, we first train a 2D diffusion prior. Given the camera pose in novel views, our approach transfers the camera information into explicit hand representations, including hand depth and skeleton images. We propose a global hand embedding to control the diffusion model based on these hand representations. We then learn a 3D Gaussian splatting for novel-view rendering using the diffusion prior. However, occluded objects present a persistent challenge. To address this issue, we further introduce local hand embedding, where a contact field is defined in the 3D Gaussian Splatting. We leverage contact information to guide the rendering in the contact field. Extensive experiments on the HO3D and DexYCB datasets demonstrate that our method significantly outperforms state-of-the-art novel-view synthesis for hand-object interactions.

Real-world datasets follow an imbalanced distribution, which poses significant challenges in rare-category object detection. Recent studies tackle this problem by developing re-weighting and re-sampling methods, that utilise the class frequencies of the dataset. However, these techniques focus solely on the frequency statistics and ignore the distribution of the classes in image space, missing important information. In contrast to them, we propose FRActal CALibration (FRACAL): a novel post-calibration method for long-tailed object detection. FRACAL devises a logit adjustment method that utilises the fractal dimension to estimate how uniformly classes are distributed in image space. During inference, it uses the fractal dimension to inversely down-weight the probabilities of uniformly spaced class predictions achieving balance in two axes: between frequent and rare categories, and between uniformly spaced and sparsely spaced classes. FRACAL is a post-processing method and it does not require any training, also it can be combined with many off-the-shelf models such as one-stage sigmoid detectors and two-stage instance segmentation models. FRACAL boosts the rare class performance by up to 8.6% and surpasses all previous methods on LVIS dataset, while showing good generalisation to other datasets such as COCO, V3Det and OpenImages. We provide the code at https://github.com/kostas1515/FRACAL.

In recent years, 3D hand pose estimation methods have garnered significant attention due to their extensive applications in human-computer interaction, virtual reality, and robotics. In contrast, there has been a notable gap in hand detection pipelines, posing significant challenges in constructing effective real-world multi-hand reconstruction systems. In this work, we present a data-driven pipeline for efficient multi-hand reconstruction in the wild. The proposed pipeline is composed of two components: a real-time fully convolutional hand localization and a high-fidelity transformer-based 3D hand reconstruction model. To tackle the limitations of previous methods and build a robust and stable detection network, we introduce a large-scale dataset with over than 2M in-the-wild hand images with diverse lighting, illumination, and occlusion conditions. Our approach outperforms previous methods in both efficiency and accuracy on popular 2D and 3D benchmarks. Finally, we showcase the effectiveness of our pipeline to achieve smooth 3D hand tracking from monocular videos, without utilizing any temporal components. Code, models, and dataset are available on our project page.

We present CaricatureBooth, a system that transforms caricature creation into a simple interactive experience – as easy as using a photo booth! A key challenge in caricature generation is two-fold: the scarcity of high-quality caricature data and the difficulty in enabling precise creative control over the exaggeration process while maintaining identity. Prior approaches either require large-scale caricature and photo data or lack intuitive mechanisms for users to guide the deformation without losing identity. We address the data scarcity by synthesising training data through Thin Plate Spline (TPS) deformation of standard face images. For creative control, we design a Bézier curve interface where users can easily manipulate facial features, with these edits then driving TPS transformations at inference time. When combined with a pre-trained ID-preserving diffusion model, our system maintains both identity preservation and creative flexibility. Through extensive experiments, we demonstrate that CaricatureBooth achieves state-of-the-art quality while making the joy of caricature creation as accessible as taking a photo – just walk in and walk out with your personalised caricature! Code is available at https://github.com/WinKawaks/CaricatureBooth.

Recent 3D face reconstruction methods have made remarkable advancements, yet achieving high-quality facial reflectance from monocular input remains challenging. Existing methods rely on the light-stage captured data to learn facial reflectance models. However, limited subject diversity in these datasets poses challenges in achieving good generalization and broad applicability. This motivates us to explore whether the extensive priors captured in recent generative diffusion models (e.g., Stable Diffusion) can enable more generalizable facial reflectance estimation as these models have been pre-trained on large-scale internet image collections containing rich visual patterns. In this paper, we introduce the use of Stable Diffusion as a prior for facial reflectance estimation, achieving robust results with minimal captured data for fine-tuning. We present S3-Face, a comprehensive framework capable of producing SSS-compliant skin reflectance from in-the-wild images. Our method adopts a two-stage training approach: in the first stage, DSN-Net is trained to predict diffuse albedo, specular albedo, and normal maps from in-the-wild images using a novel joint reflectance attention module. In the second stage, HM-Net is trained to generate hemoglobin and melanin maps based on the diffuse albedo predicted in the first stage, yielding SSS-compliant and detailed reflectance maps. Extensive experiments demonstrate that our method achieves strong generalization and produces high-fidelity, SSS-compliant facial reflectance estimation.

Despite the advent in 3D hand pose estimation, current methods predominantly focus on single-image 3D hand reconstruction in the camera frame, overlooking the world-space motion of the hands. Such limitation prohibits their direct use in egocentric video settings, where hands and camera are continuously in motion. In this work, we propose HaWoR, a high-fidelity method for hand motion reconstruction in world coordinates from egocentric videos. We propose to decouple the task by reconstructing the hand motion in the camera space and estimating the camera trajectory in the world coordinate system. To achieve precise camera trajectory estimation, we propose an adaptive egocentric SLAM framework that addresses the shortcomings of traditional SLAM methods, providing robust performance under challenging camera dynamics. To ensure robust hand motion trajectories, even when the hands move out of view frustum, we devise a novel motion infiller net work that effectively completes the missing frames of the sequence. Through extensive quantitative and qualitative evaluations, we demonstrate that HaWoR achieves state-of-the-art performance on both hand motion reconstruction and world-frame camera trajectory estimation under different egocentric benchmark datasets. Code and models are available on https://hawor-project.github.io/.

Adversarial attack severely degrades the performance of deep face recognition models with minimal image perturbations. Facial attacks can be broadly categorized into dodging and impersonation. Although existing approaches have made satisfactory progress by separately addressing each type of attack, the unification of them still remains unexplored and the potential benefit might be ignored. In this work, we aim to construct a unified facial attack framework that can simultaneously perform both attacks. We first design Unified Attack Modeling, which aligns optimization directions and spaces of two attacks to harmonize optimizing forces of dodging and impersonation. Based on the unified optimization goals, we introduce an Attacking by Decomposing scheme. The proposed scheme leverages the unified optimization goal by decomposing adversarial perturbation synthesis into two steps, orientation exploration and amplified forwarding. Extensive experiments on several facial datasets LFW, CelebA-HQ, and FFHQ show that our proposed unified framework achieves significant performance improvement on dodging and impersonation attacks, compared with the state-of-the-art methods.

Although diffusion prior is rising as a powerful solution for blind face restoration (BFR), the inherent gap between the vanilla diffusion model and BFR settings hinders its seamless adaptation. The gap mainly stems from the discrepancy between 1) high-quality (HQ) and low-quality (LQ) images and 2) synthesized and real-world images. The vanilla diffusion model is trained on images with no or less degradations, whereas BFR handles moderately to severely degraded images. Additionally, LQ images used for training are synthesized by a naive degradation model with limited degradation patterns, which fails to simulate complex and unknown degradations in real-world scenarios. In this work, we use a unified network FLIPNET that switches between two modes to resolve specific gaps. In Restoration mode, the model gradually integrates BFR-oriented features and face embeddings from LQ images to achieve authentic and faithful face restoration. In Degradation mode, the model synthesizes real-world like degraded images based on the knowledge learned from real-world degradation datasets. Extensive evaluations on benchmark datasets show that our model 1) outperforms previous diffusion prior based BFR methods in terms of authenticity and fidelity, and 2) outperforms the naive degradation model in modeling the real-world degradations.

Over the last years, 3D morphable models (3DMMs) have emerged as a state-of-the-art methodology for modeling and generating expressive 3D avatars. However, given their reliance on a strict topology, along with their linear nature, they struggle to represent complex full-head shapes. Following the advent of deep implicit functions, we propose imHead, a novel implicit 3DMM that not only models expressive 3D head avatars but also facilitates localized editing of the facial features. Previous methods directly divided the latent space into local components accompanied by an identity encoding to capture the global shape variations, leading to expensive latent sizes. In contrast, we retain a single compact identity space and introduce an intermediate region-specific latent representation to enable local edits. To train imHead, we curate a large-scale dataset of 4K distinct identities, making a step-towards large scale 3D head modeling. Under a series of experiments we demonstrate the expressive power of the proposed model to represent diverse identities and expressions outperforming previous approaches. Additionally, the proposed approach provides an interpretable solution for 3D face manipulation, allowing the user to make localized edits.

Learning visual representations is foundational for a broad spectrum of downstream tasks. Although recent vision-language contrastive models, such as CLIP and SigLIP, have achieved impressive zero-shot performance via large-scale vision-language alignment, their reliance on global representations constrains their effectiveness for dense prediction tasks, such as grounding, OCR, and segmentation. To address this gap, we introduce Region-Aware Cluster Discrimination (RICE), a novel method that enhances region-level visual and OCR capabilities. We first construct a billion-scale candidate region dataset and propose a Region Transformer layer to extract rich regional semantics. We further design a unified region cluster discrimination loss that jointly supports object and OCR learning within a single classification framework, enabling efficient and scalable distributed training on large-scale data. Extensive experiments show that RICE consistently outperforms previous methods on tasks, including segmentation, dense detection, and visual perception for Multimodal Large Language Models (MLLMs). The pre-trained models have been released at https://github.com/deepglint/MVT.

We study two strange phenomena in auto-regressive Transformers: (1) the dominance of the first token in attention heads; (2) the occurrence of large outlier activations in the hidden states. We find that popular large language models, such as Llama attend maximally to the first token in 98% of attention heads, a behaviour we attribute to the softmax function. To mitigate this issue, we propose a reformulation of softmax to softmax-1. Furthermore, we identify adaptive optimisers, e.g. Adam, as the primary contributor to the large outlier activations and introduce OrthoAdam, a novel optimiser that utilises orthogonal matrices to transform gradients, to address this issue. Finally, not only do our methods prevent these phenomena from occurring, but additionally, they enable Transformers to sustain their performance when quantised using basic algorithms, something that standard methods are unable to do. In summary, our methods reduce the attention proportion on the first token from 65% to 3.3%, the activation kurtosis in the hidden states from 1657 to 3.1, and perplexity penalty under 4-bit weight quantisation from 3565 to 0.3. Code is available at https://github.com/prannaykaul/OrthoAdam

The Contrastive Language-Image Pre-training (CLIP) framework has become a widely used approach for multimodal representation learning, particularly in image-text retrieval and clustering. However, its efficacy is constrained by three key limitations: (1) text token truncation, (2) isolated image-text encoding, and (3) deficient compositionality due to bag-of-words behavior. While recent Multimodal Large Language Models (MLLMs) have demonstrated significant advances in generalized vision-language understanding, their potential for learning transferable multimodal representations remains underexplored. In this work, we present UniME (Universal Multimodal Embedding), a novel two-stage framework that leverages MLLMs to learn discriminative representations for diverse downstream tasks. In the first stage, we perform textual discriminative knowledge distillation from a powerful LLM-based teacher model to enhance the embedding capability of the MLLM's language component. In the second stage, we introduce hard negative enhanced instruction tuning to further advance discriminative representation learning. Specifically, we initially mitigate false negative contamination and then sample multiple hard negatives per instance within each batch, forcing the model to focus on challenging samples. This approach not only improves discriminative power but also enhances instruction-following ability in downstream tasks. We conduct extensive experiments on the MMEB benchmark and multiple retrieval tasks, including short & long caption retrieval and compositional retrieval. Results demonstrate that UniME achieves consistent performance improvement across all tasks, exhibiting superior discriminative and compositional capabilities. The code will be released in https://garygutc.github.io/UniME.

After pre-training on extensive image-text pairs, Contrastive Language-Image Pre-training (CLIP) demonstrates promising performance on a wide variety of benchmarks. However, a substantial volume of multimodal interleaved documents remains underutilized for contrastive vision-language representation learning. To fully leverage these unpaired documents, we initially establish a Real-World Data Extraction pipeline to extract high-quality images and texts. Then we design a hierarchical retrieval method to efficiently associate each image with multiple semantically relevant realistic texts. To further enhance fine-grained visual information, we propose an image semantic augmented generation module for synthetic text production. Furthermore, we employ a semantic balance sampling strategy to improve dataset diversity, enabling better learning of long-tail concepts. Based on these innovations, we construct RealSyn, a dataset combining realistic and synthetic texts, available in three scales: 15M, 30M, and 100M. We compare our dataset with other widely used datasets of equivalent scale for CLIP training. Models pre-trained on RealSyn consistently achieve state-of-the-art performance across various downstream tasks, including linear probe, zero-shot transfer, zero-shot robustness, and zero-shot retrieval. Furthermore, extensive experiments confirm that RealSyn significantly enhances contrastive vision-language representation learning and demonstrates robust scalability. The code will be released in https://garygutc.github.io/RealSyn.

Neural radiance fields (NeRF) and 3D Gaussian Splatting (3DGS) are popular techniques to reconstruct and render photorealistic images. However, the prerequisite of running Structure-from-Motion (SfM) to get camera poses limits their completeness. Although previous methods can reconstruct a few unposed images, they are not applicable when images are unordered or densely captured. In this work, we propose a method to train 3DGS from unposed images. Our method leverages a pre-trained 3D geometric foundation model as the neural scene representation. Since the accuracy of the predicted pointmaps does not suffice for accurate image registration and high-fidelity image rendering, we propose to mitigate the issue by initializing and fine-tuning the pre-trained model from a seed image. The images are then progressively registered and added to the training buffer, which is used to train the model further. We also propose to refine the camera poses and pointmaps by minimizing a point-to-camera ray consistency loss across multiple views. When evaluated on diverse challenging datasets, our method outperforms state-of-the-art pose-free NeRF/3DGS methods in terms of both camera pose accuracy and novel view synthesis, and even renders higher fidelity images than 3DGS trained with COLMAP poses.

Despite the impressive progress of recent pretraining methods on multimodal tasks, existing methods are inherently biased towards either spatial modeling (e.g., CLIP) or temporal modeling (e.g., V-JEPA), limiting their joint capture of spatial details and temporal dynamics. To this end, we propose UniViT, a cluster-driven unified self-supervised learning framework that effectively captures the structured semantics of both image spatial content and video temporal dynamics through event-level and object-level clustering and discrimination. Specifically, we leverage offline clustering to generate semantic clusters across both modalities. For videos, multi-granularity event-level clustering progressively expands from single-event to structured multi-event segments, capturing coarse-to-fine temporal semantics; for images, object-level clustering captures fine-grained spatial semantics. However, while global clustering provides semantically consistent clusters, it lacks modeling of structured semantic relations (e.g., temporal event structures). To address this, we introduce a contrastive objective that leverages these semantic clusters as pseudo-label supervision to explicitly enforce structural constraints, including temporal event relations and spatial object co-occurrences, capturing structured semantics beyond categories. Meanwhile, UniViT jointly embeds structured object-level and event-level semantics into a unified representation space. Furthermore, UniViT introduces two key components: (i) Unified Rotary Position Embedding integrates relative positional embedding with frequency-aware dimension allocation to support position-invariant semantic learning and enhance the stability of structured semantics in the discrimination stage; and (ii) Variable Spatiotemporal Streams adapt to inputs of varying frame lengths, addressing the rigidity of conventional fixed-input approaches. Extensive experiments across varying model scales demonstrate that UniViT achieves state-of-the-art performance on linear probing, attentive probing, question answering, and spatial understanding tasks.

3D generation from natural language offers significant potential to reduce expert manual modeling efforts and enhance accessibility to 3D assets. However, existing methods often yield unstructured meshes and exhibit poor interactivity, making them impractical for artistic workflows. To address these limitations, we represent 3D assets as shape programs and introduce ShapeCraft, a novel multi-agent framework for text-to-3D generation. At its core, we propose a Graph-based Procedural Shape (GPS) representation that decomposes complex natural language into a structured graph of sub-tasks, thereby facilitating accurate LLM comprehension and interpretation of spatial relationships and semantic shape details. Specifically, LLM agents hierarchically parse user input to initialize GPS, then iteratively refine procedural modeling and painting to produce structured, textured, and interactive 3D assets. Qualitative and quantitative experiments demonstrate ShapeCraft's superior performance in generating geometrically accurate and semantically rich 3D assets compared to existing LLM-based agents. We further show the versatility of ShapeCraft through examples of animated and user-customized editing, highlighting its potential for broader interactive applications.

End-to-end autonomous driving methods aim to directly map raw sensor inputs to future driving actions such as planned trajectories, bypassing traditional modular pipelines. While these approaches have shown promise, they often operate under a one-shot paradigm that relies heavily on the current scene context, potentially underestimating the importance of scene dynamics and their temporal evolution. This limitation restricts the model’s ability to make informed and adaptive decisions in complex driving scenarios. We propose a new perspective: the future trajectory of an autonomous vehicle is closely intertwined with the evolving dynamics of its environment, and conversely, the vehicle’s own future states can influence how the surrounding scene unfolds. Motivated by this bidirectional relationship, we introduce **SeerDrive**, a novel end-to-end framework that jointly models future scene evolution and trajectory planning in a closed-loop manner. Our method first predicts future bird’s-eye view (BEV) representations to anticipate the dynamics of the surrounding scene, then leverages this foresight to generate future-context-aware trajectories. Two key components enable this: (1) future-aware planning, which injects predicted BEV features into the trajectory planner, and (2) iterative scene modeling and vehicle planning, which refines both future scene prediction and trajectory generation through collaborative optimization. Extensive experiments on the NAVSIM and nuScenes benchmarks show that SeerDrive significantly outperforms existing state-of-the-art methods.

Gait benchmarks empower the research community to train and evaluate high-performance gait recognition systems. Even though growing efforts have been devoted to cross-view recognition, academia is restricted by current existing databases captured in the controlled environment. In this paper, we contribute a new benchmark and strong baseline for Gait REcognition in the Wild (GREW). The GREW dataset is constructed from natural videos, which contain hundreds of cameras and thousands of hours of streams in open systems. With tremendous manual annotations, the GREW consists of 26 K identities and 128 K sequences with rich attributes for unconstrained gait recognition. Moreover, we add a distractor set of over 233 K sequences, making it more suitable for real-world applications. Compared with prevailing predefined cross-view datasets, the GREW has diverse and practical view variations, as well as more naturally challenging factors. To the best of our knowledge, this is the first large-scale dataset for gait recognition in the wild. Equipped with this benchmark, we dissect the unconstrained gait recognition problem, where representative appearance-based and model-based methods are explored. The proposed GREW benchmark proves to be essential for both training and evaluating gait recognizers in unconstrained scenarios. In addition, we propose the Single Path One-Shot neural architecture search with uniform sampling for Gait recognition, named SPOSGait, which is the first NAS-based gait recognition model. In experiments, SPOSGait achieves state-of-the-art performance on the CASIA-B, OU-MVLP, Gait3D, and GREW benchmarks, outperforming existing approaches by a large margin.

Sound Event Detection (SED) aims to predict the temporal boundaries of all the events of interest and their class labels, given an unconstrained audio sample. Taking either the split-and-classify (i.e., frame-level) strategy or the more principled event-level modeling approach, all existing methods consider the SED problem from the discriminative learning perspective. In this work, we reformulate the SED problem by taking a generative learning perspective. Specifically, we aim to generate sound temporal boundaries from noisy proposals in a denoising diffusion process, conditioned on a target audio sample. During training, our model learns to reverse the noising process by converting noisy latent queries to the ground-truth versions in the elegant Transformer decoder framework. Doing so enables the model generate accurate event boundaries from even noisy queries during inference. Extensive experiments on the Urban-SED and EPIC-Sounds datasets demonstrate that our model significantly outperforms existing alternatives, with 40+% faster convergence in training. Code: https://github.com/Surrey-UPLab/DiffSED

We address the challenging problem of Long-Tailed Semi-Supervised Learning (LTSSL) where labeled data exhibit imbalanced class distribution and unlabeled data follow an unknown distribution. Unlike in balanced SSL, the generated pseudo-labels are skewed towards head classes, intensifying the training bias. Such a phenomenon is even amplified as more unlabeled data will be mislabeled as head classes when the class distribution of labeled and unlabeled datasets are mismatched. To solve this problem, we propose a novel method named ComPlementary Experts (CPE). Specifically, we train multiple experts to model various class distributions, each of them yielding high-quality pseudo-labels within one form of class distribution. Besides, we introduce Classwise Batch Normalization for CPE to avoid performance degradation caused by feature distribution mismatch between head and non-head classes. CPE achieves state-of-the-art performances on CIFAR-10-LT, CIFAR-100-LT, and STL-10-LT dataset benchmarks. For instance, on CIFAR-10-LT, CPE improves test accuracy by over >2.22% compared to baselines. Code is available at https://github.com/machengcheng2016/CPE-LTSSL.

Stable diffusion, a generative model used in text-to-image synthesis, frequently encounters resolution-induced composition problems when generating images of varying sizes. This issue primarily stems from the model being trained on pairs of single-scale images and their corresponding text descriptions. Moreover, direct training on images of unlimited sizes is unfeasible, as it would require an immense number of text-image pairs and entail substantial computational expenses. To overcome these challenges, we propose a two-stage pipeline named Any-Size-Diffusion (ASD), designed to efficiently generate well-composed HD images of any size, while minimizing the need for high-memory GPU resources. Specifically, the initial stage, dubbed Any Ratio Adaptability Diffusion (ARAD), leverages a selected set of images with a restricted range of ratios to optimize the text-conditional diffusion model, thereby improving its ability to adjust composition to accommodate diverse image sizes. To support the creation of images at any desired size, we further introduce a technique called Fast Seamless Tiled Diffusion (FSTD) at the subsequent stage. This method allows for the rapid enlargement of the ASD output to any high-resolution size, avoiding seaming artifacts or memory overloads. Experimental results on the LAION-COCO and MM-CelebA-HQ benchmarks demonstrate that ASD can produce well-structured images of arbitrary sizes, cutting down the inference time by 2X compared to the traditional tiled algorithm. The source code is available at https://github.com/ProAirVerse/Any-Size-Diffusion.

Sign Languages (SL) serve as the primary mode of communication for the Deaf and Hard of Hearing communities. Deep learning methods for SL recognition and translation have achieved promising results. However, Sign Language Production (SLP) poses a challenge as the generated motions must be realistic and have precise semantic meaning. Most SLP methods rely on 2D data, which hinders their re-alism. In this work, a diffusion-based SLP model is trained on a curated large-scale dataset of 4D signing avatars and their corresponding text transcripts. The proposed method can generate dynamic sequences of 3D avatars from an un-constrained domain of discourse using a diffusion process formed on a novel and anatomically informed graph neu-ral network defined on the SMPL-X body skeleton. Through quantitative and qualitative experiments, we show that the proposed method considerably outperforms previous meth-ods of SLP. This work makes an important step towards re-alistic neural sign avatars, bridging the communication gap between Deaf and hearing communities.11Project page: https://baltatzisv.github.io/neural-sign-actors/

Detecting objects in low-light scenarios presents a persistent challenge, as detectors trained on well-lit data exhibit significant performance degradation on low-light data due to low visibility. Previous methods mitigate this issue by exploring image enhancement or object detection techniques with real low-light image datasets. However, the progress is impeded by the inherent difficulties about collecting and annotating low-light images. To address this challenge, we propose to boost low-light object detection with zero-shot day-night domain adaptation, which aims to generalize a detector from well-lit scenarios to low-light ones without requiring real low-light data. Re-visiting Retinex theory in the low-level vision, we first de-sign a reflectance representation learning module to learn Retinex-based illumination invariance in images with a carefully designed illumination invariance reinforcement strategy. Next, an interchange-redecomposition-coherence procedure is introduced to improve over the vanilla Retinex image decomposition process by performing two sequential image decompositions and introducing a redecomposition cohering loss. Extensive experiments on ExDark, DARK FACE, and CODaN datasets show strong low-light generalizability of our method. Our code is available at https://github.com/ZPDu/DAI-Net.

Domain generalization (DG) based Face Anti-Spoofing (FAS) aims to improve the model's performance on unseen domains. Existing methods either rely on domain labels to align domain-invariant feature spaces, or disentangle generalizable features from the whole sample, which inevitably lead to the distortion of semantic feature structures and achieve limited generalization. In this work, we make use of large-scale VLMs like CLIP and leverage the textual feature to dynamically adjust the classifier's weights for exploring generalizable visual features. Specifically, we propose a novel Class Free Prompt Learning (CFPL) paradigm for DG FAS, which utilizes two lightweight transformers, namely Content Q-Former (CQF) and Style Q-Former (SQF), to learn the different semantic prompts conditioned on content and style features by using a set of learnable query vectors, respectively. Thus, the generalizable prompt can be learned by two improvements: (1) A Prompt-Text Matched (PTM) supervision is introduced to ensure CQF learns visual representation that is most informative of the content description. (2) A Diversified Style Prompt (DSP) technology is proposed to diversify the learning of style prompts by mixing feature statistics between instance-specific styles. Finally, the learned text features modulate visual features to generalization through the designed Prompt Modulation (PM). Extensive experiments show that the CFPL is effective and outperforms the state-of-the-art methods on several cross-domain datasets.

Although diffusion models are rising as a powerful solution for blind face restoration, they are criticized for two problems: 1) slow training and inference speed, and 2)failure in preserving identity and recovering fine-grained facial details. In this work, we propose WaveFace to solve the problems in the frequency domain, where low- and high-frequency components decomposed by wavelet transformation are considered individually to maximize authenticity as well as efficiency. The diffusion model is applied to recover the low-frequency component only, which presents general information of the original image but 1/16 in size. To preserve the original identity, the generation is conditioned on the low-frequency component of low-quality images at each denoising step. Meanwhile, high-frequency components at multiple decomposition levels are handled by a unified network, which recovers complex facial details in a single step. Evaluations on four benchmark datasets show that: 1) WaveFace outperforms state-of-the-art methods in authenticity, especially in terms of identity preservation, and 2) authentic images are restored with the efficiency 10x faster than existing diffusion model-based BFR methods.

Knowledge distillation is an effective method for training small and efficient deep learning models. However, the efficacy of a single method can degenerate when transferring to other tasks, modalities, or even other architectures. To address this limitation, we propose a novel constrained feature distillation method. This method is derived from a small set of core principles, which results in two emerging components: an orthogonal projection and a task-specific normalisation. Equipped with both of these components, our transformer models can outperform all previous methods on ImageNet and reach up to a 4.4% relative improvement over the previous state-of-the-art methods. To further demonstrate the generality of our method, we apply it to object detection and image generation, whereby we obtain consistent and substantial performance improvements over state-of-the-art. Code and models are publicly available11https://github.com/roymiles/vkd.

We introduce a novel 3D generative method, Generative 3D Reconstruction (G3DR) in ImageNet, capable of generating diverse and high-quality 3D objects from single images, addressing the limitations of existing methods. At the heart of our framework is a novel depth reg-ularization technique that enables the generation of scenes with high-geometric fidelity. G3DR also leverages a pre-trained language-vision model, such as CLIP, to enable reconstruction in novel views and improve the visual realism of generations. Additionally, G3DR designs a simple but effective sampling procedure to further improve the quality of generations. G3DR offers diverse and efficient 3D asset generation based on class or text conditioning. Despite its simplicity, G3DR is able to beat state-of-the-art methods, improving over them by up to 22% in per-ceptual metrics and 90% in geometry scores, while needing only half of the training time. Code is available at https://github.com/preddy5/G3DR.

Recent 3D face reconstruction methods have made re-markable advancements, yet there remain huge challenges in monocular high-quality facial reflectance reconstruction. Existing methods rely on a large amount of light-stage captured data to learn facial reflectance models. However, the lack of subject diversity poses challenges in achieving good generalization and widespread applicability. In this paper, we learn the reflectance prior in image space rather than UV space and present a framework named ID2Reflectance. Our framework can directly estimate the reflectance maps of a single image while using limited reflectance data for training. Our key insight is that reflectance data shares facial structures with RGB faces, which enables obtaining expressive facial prior from inexpensive RGB data thus re-ducing the dependency on reflectance data. We first learn a high-quality prior for facial reflectance. Specifically, we pretrain multi-domain facial feature code books and design a codebook fusion method to align the reflectance and RGB domains. Then, we propose an identity-conditioned swapping module that injects facial identity from the target image into the pre-trained autoencoder to modify the identity of the source reflectance image. Finally, we stitch multi-view swapped reflectance images to obtain renderable assets. Extensive experiments demonstrate that our method exhibits excellent generalization capability and achieves state-of-the-art facial reflectance reconstruction results for in-the-wild faces. Our project page is https://xingyuren.github.io/id2reflectance.

Artifact-free super-resolution (SR) aims to translate low-resolution images into their high-resolution counterparts with a strict integrity of the original content, eliminating any distortions or synthetic details. While traditional diffusion-based SR techniques have demonstrated remarkable abilities to enhance image detail, they are prone to ar-tifact introduction during iterative procedures. Such arti-facts, ranging from trivial noise to unauthentic textures, de-viate from the true structure of the source image, thus chal-lenging the integrity of the super-resolution process. In this work, we propose Self-Adaptive Reality-Guided Diffusion (SARGD), a training-free method that delves into the latent space to effectively identify and mitigate the propagation of artifacts. Our SARGD begins by using an artifact detector to identify implausible pixels, creating a binary mask that highlights artifacts. Following this, the Reality Guidance Refinement (RGR) process refines artifacts by integrating this mask with realistic latent representations, improving alignment with the original image. Nonetheless, initial realistic-latent representations from lower-quality images result in over-smoothing in the final output. To address this, we introduce a Self-Adaptive Guidance (SAG) mechanism. It dynamically computes a reality score, enhancing the sharpness of the realistic latent. These alternating mechanisms collectively achieve artifact-free super-resolution. Extensive experiments demonstrate the superiority of our method, delivering detailed artifact-free high-resolution images while reducing sampling steps by 2 x. We release our code at https://github.com/ProAirVerse/Self-Adaptive-Guidance-Diffusion.git.

Contrastive Language-Image Pre-training (CLIP) has significantly improved performance in various vision-language tasks by expanding the dataset with image-text pairs obtained from the web.This paper further explores CLIP from the perspectives of data and model architecture.To mitigate the impact of the noise data and enhance the quality of large-scale image-text data crawled from the internet, we introduce a diverse description generation framework that can leverage Large Language Models (LLMs) to combine and refine information from web-based image-text pairs, synthetic captions, and detection tags.Additionally, we propose RWKV-CLIP, the first RWKV-driven vision-language representation learning model that combines the effective parallel training of transformers with the efficient inference of RNNs.Extensive experiments across different model scales and pre-training datasets demonstrate that RWKV-CLIP is a robust vision-language representation learner and it achieves state-of-the-art performance across multiple downstream tasks, including linear probing, zero-shot classification, and zero-shot image-text retrieval.To facilitate future research, the code and pre-trained models are released at https: //github.com/deepglint/RWKV-CLIP.

The Mixture of Experts (MoE) paradigm provides a powerful way to decompose dense layers into smaller, modular computations often more amenable to human interpretation, debugging, and editability. However, a major challenge lies in the computational cost of scaling the number of experts high enough to achieve fine-grained specialization. In this paper, we propose the Multilinear Mixture of Experts (μMoE) layer to address this, focusing on vision models. μMoE layers enable scalable expert specialization by performing an implicit computation on prohibitively large weight tensors entirely in factorized form. Consequently, μMoEs (1) avoid the restrictively high inference-time costs of dense MoEs, yet (2) do not inherit the training issues of the popular sparse MoEs' discrete (non-differentiable) expert routing. We present both qualitative and quantitative evidence that scaling μMoE layers when fine-tuning foundation models for vision tasks leads to more specialized experts at the class-level, further enabling manual bias correction in CelebA attribute classification. Finally, we show qualitative results demonstrating the expert specialism achieved when pre-training large GPT2 and MLP-Mixer models with parameter-matched μMoE blocks at every layer, maintaining comparable accuracy. Our code is available at: https://github.com/james-oldfield/muMoE.

We propose ID-to-3D, a method to generate identity- and text-guided 3D human heads with disentangled expressions, starting from even a single casually captured ‘in-the-wild’ image of a subject. The foundation of our approach is anchored in compositionality, alongside the use of task-specific 2D diffusion models as priors for optimization. First, we extend a foundational model with a lightweight expression-aware and ID-aware architecture, and create 2D priors for geometric and texture generation, via fine-tuning only 0.2% of its available training parameters. Then, we jointly leverage a neural parametric representation for the expression of each subject and a multi-stage generation of highly detailed geometry and albedo texture. This combination of strong face identity embeddings and our neural representation enables accurate reconstruction of not only facial features but also accessories and hair, and can be meshed to provide render-ready assets for gaming and telepresence. Our results achieve an unprecedented level of id-consistent and high-quality texture and geometry generation, generalizing to a ‘world’ of unseen 3D identities, without relying on large 3D captured datasets of human assets.

Novel view acoustic synthesis (NVAS) aims to render binaural audio at any target viewpoint, given a mono audio emitted by a sound source at a 3D scene. Existing methods have proposed NeRF-based implicit models to exploit visual cues as a condition for synthesizing binaural audio. However, in addition to low efficiency originating from heavy NeRF rendering, these methods all have a limited ability of characterizing the entire scene environment such as room geometry, material properties, and the spatial relation between the listener and sound source. To address these issues, we propose a novel Audio-Visual Gaussian Splatting (AV-GS) model. To obtain a material-aware and geometry-aware condition for audio synthesis, we learn an explicit point-based scene representation with audio-guidance parameters on locally initialized Gaussian points, taking into account the space relation from the listener and sound source. To make the visual scene model audio adaptive, we propose a point densification and pruning strategy to optimally distribute the Gaussian points, with the per-point contribution in sound propagation (e.g., more points needed for texture-less wall surfaces as they affect sound path diversion). Extensive experiments validate the superiority of our AV-GS over existing alternatives on the real-world RWAS and simulation-based SoundSpaces datasets. Project page: \url{https://surrey-uplab.github.io/research/avgs/}

The field of face recognition (FR) has undergone significant advancements with the rise of deep learning. Recently, the success of unsupervised learning and graph neural networks has demonstrated the effectiveness of data structure information. Considering that the FR task can leverage large-scale training data, which intrinsically contains significant structure information, we aim to investigate how to encode such critical structure information into the latent space. As revealed from our observations, directly aligning the structure information between the input and latent spaces inevitably suffers from an overfitting problem, leading to a structure collapse phenomenon in the latent space. To address this problem, we propose TopoFR, a novel FR model that leverages a topological structure alignment strategy called PTSA and a hard sample mining strategy named SDE. Concretely, PTSA uses persistent homology to align the topological structures of the input and latent spaces, effectively preserving the structure information and improving the generalization performance of FR model. To mitigate the impact of hard samples on the latent space structure, SDE accurately identifies hard samples by automatically computing structure damage score (SDS) for each sample, and directs the model to prioritize optimizing these samples. Experimental results on popular face benchmarks demonstrate the superiority of our TopoFR over the state-of-the-art methods. Code and models are available at: https://github.com/modelscope/facechain/tree/main/face_module/TopoFR.

Large language models (LLMs) have recently emerged as powerful tools for tackling many language-processing tasks. Despite their success, training and fine-tuning these models is still far too computationally and memory intensive. In this paper, we identify and characterise the important components needed for effective model convergence using gradient descent. In doing so we find that the intermediate activations used to implement backpropagation can be excessively compressed without incurring any degradation in performance. This result leads us to a cheap and memory-efficient algorithm for both fine-tuning and pre-training LLMs. The proposed algorithm simply divides the tokens up into smaller sub-tokens before projecting them onto a fixed 1-dimensional subspace during the forward pass. These features are then coarsely reconstructed during the backward pass to implement the update rules. We confirm the effectiveness of our algorithm as being complimentary to many state-of-the-art PEFT methods on the VTAB-1k fine-tuning benchmark. Furthermore, we outperform QLoRA for fine-tuning LLaMA and show competitive performance against other memory-efficient pre-training methods on the large-scale C4 dataset.

Learning from crowds describes that the annotations of training data are obtained with crowd-sourcing services. Multiple annotators each complete their own small part of the annotations, where labeling mistakes that depend on annotators occur frequently. Modeling the label-noise generation process by the noise transition matrix is a powerful tool to tackle the label noise. In real-world crowd-sourcing scenarios, noise transition matrices are both annotator- and instance-dependent. However, due to the high complexity of annotator- and instance-dependent transition matrices (AIDTM), annotation sparsity, which means each annotator only labels a tiny part of instances, makes modeling AIDTM very challenging. Without prior knowledge, existing works simplify the problem by assuming the transition matrix is instance-independent or using simple parametric ways, which lose modeling generality. Motivated by this, we target a more realistic problem, estimating general AIDTM in practice. Without losing modeling generality, we parameterize AIDTM with deep neural networks. To alleviate the modeling challenge, we suppose every annotator shares its noise pattern with similar annotators, and estimate AIDTM via knowledge transfer. We hence first model the mixture of noise patterns by all annotators, and then transfer this modeling to individual annotators. Furthermore, considering that the transfer from the mixture of noise patterns to individuals may cause two annotators with highly different noise generations to perturb each other, we employ the knowledge transfer between identified neighboring annotators to calibrate the modeling. Theoretical analyses are derived to demonstrate that both the knowledge transfer from global to individuals and the knowledge transfer between neighboring individuals can effectively help mitigate the challenge of modeling general AIDTM. Experiments confirm the superiority of the proposed approach on synthetic and real-world crowd-sourcing data.

Unsupervised person re-identification (ReID) aims to train a feature extractor for identity retrieval without exploiting identity labels. Due to the no-reference trust in imperfect clustering results, the learning is inevitably misled by unreliable pseudo labels. Albeit the pseudo label refinement has been investigated by previous works, they generally leverage auxiliary information such as camera IDs and body part predictions. This work explores the internal characteristics of clusters to refine pseudo labels. To this end, Confidence-Guided Centroids (CGC) are proposed to provide reliable cluster-wise prototypes for feature learning. Since samples with high confidence are exclusively involved in the formation of centroids, the identity information of low-confidence samples, i.e., boundary samples, are NOT likely to contribute to the corresponding centroid. Given the new centroids, the current learning scheme, where samples are forced to learn from their assigned centroids solely, is unwise. To remedy the situation, we propose to use Confidence-Guided pseudo Label (CGL), which enables samples to approach not only the originally assigned centroid but also other centroids that are potentially embedded with their identity information. Empowered by confidence-guided centroids and labels, our method yields comparable performance with, or even outperforms, state-of-the-art pseudo label refinement works that largely leverage auxiliary information.

Domain shift across crowd data severely hinders crowd counting models to generalize to unseen scenarios. Although domain adaptive crowd counting approaches close this gap to a certain extent, they are still dependent on the target domain data to adapt (e.g. finetune) their models to the specific domain. In this paper, we instead target to train a model based on a single source domain which can generalize well on any unseen domain. This falls into the realm of domain generalization that remains unexplored in crowd counting. We first introduce a dynamic sub-domain division scheme which divides the source domain into multiple sub-domains such that we can initiate a meta-learning framework for domain generalization. The sub-domain division is dynamically refined during the meta-learning. Next, in order to disentangle domain-invariant information from domain-specific information in image features, we design the domain-invariant and -specific crowd memory modules to re-encode image features. Two types of losses, i.e. feature reconstruction and orthogonal losses, are devised to enable this disentanglement. Extensive experiments on several standard crowd counting benchmarks i.e. SHA, SHB, QNRF, and NWPU, show the strong generalizability of our method. Our code is available at: https://github.com/ZPDu/Domain-general-Crowd-Counting-in-Unseen-Scenarios

Deep Neural Networks (DNNs) have obtained impressive performance across tasks, however they still remain as black boxes, e.g., hard to theoretically analyze. At the same time, Polynomial Networks (PNs) have emerged as an alternative method with a promising performance and improved interpretability but have yet to reach the performance of the powerful DNN baselines. In this work, we aim to close this performance gap. We introduce a class of PNs, which are able to reach the performance of ResNet across a range of six benchmarks. We demonstrate that strong regularization is critical and conduct an extensive study of the exact regularization schemes required to match performance. To further motivate the regularization schemes, we introduce D-PolyNets that achieve a higher- degree of expansion than previously proposed polynomial networks. D-PolyNets are more parameter-efficient while achieving a similar performance as other polynomial networks. We expect that our new models can lead to an understanding of the role of elementwise activation functions (which are no longer required for training PNs). The source code is available at https://github.com/grigorisg9gr/regularized_polynomials.

In this paper, we introduce FitMe, a facial reflectance model and a differentiable rendering optimization pipeline, that can be used to acquire high-fidelity renderable human avatars from single or multiple images. The model consists of a multi-modal style-based generator, that captures facial appearance in terms of diffuse and specular reflectance, and a PCA-based shape model. We employ a fast differentiable rendering process that can be used in an optimization pipeline, while also achieving photorealistic facial shading. Our optimization process accurately captures both the facial reflectance and shape in high-detail, by exploiting the expressivity of the style-based latent representation and of our shape model. FitMe achieves state-of-the-art reflectance acquisition and identity preservation on single “in-the-wild” facial images, while it produces impressive scan-like results, when given multiple unconstrained facial images pertaining to the same identity. In contrast with recent implicit avatar reconstructions, FitMe requires only one minute and produces relightable mesh and texture-based avatars, that can be used by end-user applications.

Recent 3D face reconstruction methods have made significant advances in geometry prediction, yet further cosmetic improvements are limited by lagged albedo because inferring albedo from appearance is an ill-posed problem. Although some existing methods consider prior knowledge from illumination to improve albedo estimation, they still produce a light-skin bias due to racially biased albedo models and limited light constraints. In this paper, we reconsider the relationship between albedo and face attributes and propose a ID2Albedo to directly estimate albedo without constraining illumination. Our key insight is that intrinsic semantic attributes such as race, skin color, and age can be used to constrain the albedo map. We first introduce visual-textual cues and design a semantic loss to supervise facial albedo estimation. Specifically, we pre-define text labels such as race, skin color, age, and wrinkles. Then, we employ the text-image model (CLIP) to compute the similarity between the text and the input image, and assign a pseudo-label to each facial image. We constrain generated albedos in the training phase to have the same attributes as the inputs. In addition, we train a high-quality, unbiased facial albedo generator and utilize the semantic loss to learn the mapping from illumination-robust identity features to the albedo latent codes. Finally, our ID2Albedo is trained in a self-supervised way and outperforms state-of-the-art albedo estimation methods in terms of accuracy and fidelity. It is worth mentioning that our approach has excellent generalizability and fairness, especially on in-the- wild data.

Learning a dense 3D model with fine-scale details from a single facial image is highly challenging and ill-posed. To address this problem, many approaches fit smooth geometries through facial prior while learning details as additional displacement maps or personalized basis. However, these techniques typically require vast datasets of paired multi-view data or 3D scans, whereas such datasets are scarce and expensive. To alleviate heavy data dependency, we present a robust texture-guided geometric detail recovery approach using only a single in-the-wild facial image. Specifically, we inpaint occluded facial parts, generate complete textures, and build an accurate multi-view dataset of the target subject. In order to estimate the detailed geometry, we define an implicit signed distance function and employ a physically-based implicit renderer to reconstruct fine geometric details from the generated multiview images. Our method not only recovers accurate facial details but also decomposes the diffuse and specular albedo, normals and shading components in a self-supervised way. Finally, we register the implicit shape details to a 3D Morphable Model template, which can be used in traditional modeling and rendering pipelines. Extensive experiments demonstrate that the proposed approach can reconstruct impressive facial details from a single image, especially when compared with state-of-the-art methods trained on large datasets.

The success of deep learning models on structured data has generated significant interest in extending their application to non-Euclidean domains. In this work, we introduce a novel intrinsic operator suitable for representation learning on 3D meshes. Our operator is specifically tailored to adapt its behavior to the irregular structure of the underlying graph and effectively utilize its long-range dependencies, while at the same time ensuring computational efficiency and ease of optimization. In particular, inspired by the framework of Spiral Convolution, which extracts and transforms the vertices in the 3D mesh following a local spiral ordering, we propose a general operator that dynamically adjusts the length of the spiral trajectory and the parameters of the transformation for each processed vertex and mesh. Then, we use polyadic decomposition to factorize its dense weight tensor into a sequence of lighter linear layers that separately process features and vertices information, hence significantly reducing the computational complexity without introducing any stringent inductive biases. Notably, we leverage dynamic gating to achieve spatial adaptivity and induce global reasoning with constant time complexity benefitting from an efficient dynamic pooling mechanism based on Summed-Area-tables. Used as a drop-in replacement on existing architectures for shape correspondence our operator significantly improves the performance-efficiency trade-off, and in 3D shape generation with morphable models achieves state-of-the-art performance with a three-fold reduction in the number of parameters required. Project page: https://github.com/Fb2221/DFC

Vision Transformers (ViTs) have demonstrated powerful representation ability in various visual tasks thanks to their intrinsic data-hungry nature. However, we unexpectedly find that ViTs perform vulnerably when applied to face recognition (FR) scenarios with extremely large datasets. We investigate the reasons for this phenomenon and discover that the existing data augmentation approach and hard sample mining strategy are incompatible with ViTs-based FR backbone due to the lack of tailored consideration on preserving face structural information and leveraging each local token information. To remedy these problems, this paper proposes a superior FR model called TransFace, which employs a patch-level data augmentation strategy named DPAP and a hard sample mining strategy named EHSM. Specially, DPAP randomly perturbs the amplitude information of dominant patches to expand sample diversity, which effectively alleviates the overfitting problem in ViTs. EHSM utilizes the information entropy in the local tokens to dynamically adjust the importance weight of easy and hard samples during training, leading to a more stable prediction. Experiments on several benchmarks demonstrate the superiority of our TransFace. Code and models are available at https://github.com/DanJun6737/TransFace.

Controllable person image synthesis aims at rendering a source image based on user-specified changes in body pose or appearance. Prior art approaches leverage pixel-level denoising diffusion models conditioned on the coarse skeleton via cross-attention. This leads to two limitations: low efficiency and inaccurate condition information. To address both issues, a novel Pose-Constrained Latent Diffusion model (PoCoLD) is introduced. Rather than using the skeleton as a sparse pose representation, we exploit DensePose which offers much richer body structure information. To effectively capitalize DensePose at a low cost, we propose an efficient pose-constrained attention module that is capable of modeling the complex interplay between appearance and pose. Extensive experiments show that our PoCoLD outperforms the state-of-the-art competitors in image synthesis fidelity. Critically, it runs 2× faster and consumes 3.6× smaller memory than the latest diffusion-model-based alternative during inference. Our code and models are available at https://github.com/BrandonHanx/PoCoLD.

We propose a new formulation of temporal action detection (TAD) with denoising diffusion, DiffTAD in short. Taking as input random temporal proposals, it can yield action proposals accurately given an untrimmed long video. This presents a generative modeling perspective, against previous discriminative learning manners. This capability is achieved by first diffusing the ground-truth proposals to random ones (i.e., the forward/noising process) and then learning to reverse the noising process (i.e., the backward/denoising process). Concretely, we establish the denoising process in the Transformer decoder (e.g., DETR) by introducing a temporal location query design with faster convergence in training. We further propose a cross-step selective conditioning algorithm for inference acceleration. Extensive evaluations on ActivityNet and THUMOS show that our DiffTAD achieves top performance compared to previous art alternatives. The code is available at https://github.com/sauradip/DiffusionTAD.

Frame quality deterioration is one of the main challenges in the field of video understanding. To compensate for the information loss caused by deteriorated frames, recent approaches exploit transformer-based integration modules to obtain spatio-temporal information. However, these integration modules are heavy and complex. Furthermore, each integration module is specifically tailored for its target task, making it difficult to generalise to multiple tasks. In this paper, we present a neat and unified framework, called Spatio-Temporal Prompting Network (STPN). It can efficiently extract robust and accurate video features by dynamically adjusting the input features in the backbone network. Specifically, STPN predicts several video prompts containing spatio-temporal information of neighbour frames. Then, these video prompts are prepended to the patch embeddings of the current frame as the updated input for video feature extraction. Moreover, STPN is easy to generalise to various video tasks because it does not contain task-specific modules. Without bells and whistles, STPN achieves state-of-the-art performance on three widely-used datasets for different video understanding tasks, i.e., ImageNetVID for video object detection, YouTubeVIS for video instance segmentation, and GOT-10k for visual object tracking. Codes are available at https://github.com/guanxiongsun/STPN

Multi-label classification aims to learn classification models from instances associated with multiple labels. It is pivotal to learn and utilize the label dependence among multiple labels in multi-label classification. As a result of today’s big and complex data, noisy labels are inevitable, making it looming to target multi-label classification with noisy labels. Although the importance of label dependence has been shown in multi-label classification with clean labels, it is challenging and hard to bring label dependence to the problem of multi-label classification with noisy labels. The issues are, that we do not understand why label dependence is helpful in the problem, and how to learn and utilize label dependence only using training data with noisy multiple labels. In this paper, we bring label dependence to tackle the problem of multi-label classification with noisy labels. Specifically, we first provide a high-level understanding of why label dependence helps distinguish the examples with clean/noisy multiple labels. Benefiting from the memorization effect in handling noisy labels, a novel algorithm is then proposed to learn the label dependence by only employing training data with noisy multiple labels, and utilize the learned dependence to help correct noisy multiple labels to clean ones. We prove that the use of label dependence could bring a higher success rate for recovering correct multiple labels. Empirical evaluations justify our claims and demonstrate the superiority of our algorithm.

Contrastive Language-Image Pre-training (CLIP) has significantly boosted the performance of various vision-language tasks by scaling up the dataset with image-text pairs collected from the web. However, the presence of intrinsic noise and unmatched image-text pairs in web data can potentially affect the performance of representation learning. To address this issue, we first utilize the OFA model to generate synthetic captions that focus on the image content. The generated captions contain complementary information that is beneficial for pre-training. Then, we propose an Adaptive Language-Image Pre-training (ALIP), a bi-path model that integrates supervision from both raw text and synthetic caption. As the core components of ALIP, the Language Consistency Gate (LCG) and Description Consistency Gate (DCG) dynamically adjust the weights of samples and image-text/caption pairs during the training process. Meanwhile, the adaptive contrastive loss can effectively reduce the impact of noise data and enhances the efficiency of pre-training data. We validate ALIP with experiments on different scales of models and pre-training datasets. Experiments results show that ALIP achieves state-of-the-art performance on multiple downstream tasks including zero-shot image-text retrieval and linear probe. To facilitate future research, the code and pre-trained models are released at https://github.com/deepglint/ALIP.

Recently, text-guided 3D generative methods have made remarkable advancements in producing high-quality textures and geometry, capitalizing on the proliferation of large vision-language and image diffusion models. However, existing methods still struggle to create high-fidelity 3D head avatars in two aspects: (1) They rely mostly on a pre-trained text-to-image diffusion model whilst missing the necessary 3D awareness and head priors. This makes them prone to inconsistency and geometric distortions in the generated avatars. (2) They fall short in fine-grained editing. This is primarily due to the inherited limitations from the pre-trained 2D image diffusion models, which become more pronounced when it comes to 3D head avatars. In this work, we address these challenges by introducing a versatile coarse-to-fine pipeline dubbed HeadSculpt for crafting (i.e., generating and editing) 3D head avatars from textual prompts. Specifically, we first equip the diffusion model with 3D awareness by leveraging landmark-based control and a learned textual embedding representing the back view appearance of heads, enabling 3D-consistent head avatar generations. We further propose a novel identity-aware editing score distillation strategy to optimize a textured mesh with a high-resolution differentiable rendering technique. This enables identity preservation while following the editing instruction. We showcase HeadSculpt's superior fidelity and editing capabilities through comprehensive experiments and comparisons with existing methods.

Self-attention mechanisms and non-local blocks have become crucial building blocks for state-of-the-art neural architectures thanks to their unparalleled ability in capturing long-range dependencies in the input. However their cost is quadratic with the number of spatial positions hence making their use impractical in many real case applications. In this work, we analyze these methods through a polynomial lens, and we show that self-attention can be seen as a special case of a 3 rd order polynomial. Within this polynomial framework, we are able to design polynomial operators capable of accessing the same data pattern of non-local and self-attention blocks while reducing the complexity from quadratic to linear. As a result, we propose two modules (Poly-NL and Poly-SA) that can be used as "drop-in" replacements for more-complex non-local and self-attention layers in state-of-the-art CNNs and ViT architectures. Our modules can achieve comparable, if not better, performance across a wide range of computer vision tasks while keeping a complexity equivalent to a standard linear layer.

The noise transition matrix T, reflecting the probabilities that true labels flip into noisy ones, is of vital importance to model label noise and build statistically consistent classifiers. The traditional transition matrix is limited to model closed-set label noise, where noisy training data have true class labels within the noisy label set. It is unfitted to employ such a transition matrix to model open-set label noise, where some true class labels are outside the noisy label set. Therefore, when considering a more realistic situation, i.e., both closed-set and open-set label noises occur, prior works will give unbelievable solutions. Besides, the traditional transition matrix is mostly limited to model instance-independent label noise, which may not perform well in practice. In this paper, we focus on learning with the mixed closed-set and open-set noisy labels. We address the aforementioned issues by extending the traditional transition matrix to be able to model mixed label noise, and further to the cluster-dependent transition matrix to better combat the instance-dependent label noise in real-world applications. We term the proposed transition matrix as the cluster-dependent extended transition matrix. An unbiased estimator (i.e., extended T-estimator) has been designed to estimate the cluster-dependent extended transition matrix by only exploiting the noisy data. Comprehensive experiments validate that our method can better cope with realistic label noise, following its more robust performance than the prior state-of-the-art label-noise learning methods.

Recent deep face hallucination methods show stunning performance in super-resolving severely degraded facial images, even surpassing human ability. However, these algorithms are mainly evaluated on non-public synthetic datasets. It is thus unclear how these algorithms perform on public face hallucination datasets. Meanwhile, most of the existing datasets do not well consider the distribution of races, which makes face hallucination methods trained on these datasets biased toward some specific races. To address the above two problems, in this paper, we build a public Ethnically Diverse Face dataset, EDFace-Celeb-1 M, and design a benchmark task for face hallucination. Our dataset includes 1.7 million photos that cover different countries, with relatively balanced race composition. To the best of our knowledge, it is the largest-scale and publicly available face hallucination dataset in the wild. Associated with this dataset, this paper also contributes various evaluation protocols and provides comprehensive analysis to benchmark the existing state-of-the-art methods. The benchmark evaluations demonstrate the performance and limitations of state-of-the-art algorithms. https://github.com/HDCVLab/EDFace-Celeb-1M.

Face benchmarks empower the research community to train and evaluate high-performance face recognition systems. In this paper, we contribute a new million-scale recognition benchmark, containing uncurated 4M identities/260M faces (WebFace260M) and cleaned 2M identities/42M faces (WebFace42M) training data, as well as an elaborately designed time-constrained evaluation protocol. First, we collect 4M name lists and download 260M faces from the Internet. Then, a Cleaning Automatically utilizing Self-Training (CAST) pipeline is devised to purify the tremendous WebFace260M, which is efficient and scalable. To the best of our knowledge, the cleaned WebFace42M is the largest public face recognition training set and we expect to close the data gap between academia and industry. Referring to practical deployments, Face Recognition Under Inference Time conStraint (FRUITS) protocol and a new test set with rich attributes are constructed. Besides, we gather a large-scale masked face sub-set for biometrics assessment under COVID-19. For a comprehensive evaluation of face matchers, three recognition tasks are performed under standard, masked and unbiased settings, respectively. Equipped with this benchmark, we delve into million-scale face recognition problems. A distributed framework is developed to train face recognition models efficiently without tampering with the performance. Enabled by WebFace42M, we reduce 40% failure rate on the challenging IJB-C set and rank 3rd among 430 entries on NIST-FRVT. Even 10% data (WebFace4M) shows superior performance compared with the public training sets. Furthermore, comprehensive baselines are established under the FRUITS-100/500/1000 milliseconds protocols. The proposed benchmark shows enormous potential on standard, masked and unbiased face recognition scenarios. Our WebFace260M website is https://www.face-benchmark.org.

Pretrained vision-language models (VLMs) such as CLIP have shown impressive generalization capability in downstream vision tasks with appropriate text prompts. Instead of designing prompts manually, Context Optimization (CoOp) has been recently proposed to learn continuous prompts using task-specific training data. Despite the performance improvements on downstream tasks, several studies have reported that CoOp suffers from the overfitting issue in two aspects: (i) the test accuracy on base classes first improves and then worsens during training; (ii) the test accuracy on novel classes keeps decreasing. However, none of the existing studies can understand and mitigate such overfitting problems. In this study, we first explore the cause of overfitting by analyzing the gradient flow. Comparative experiments reveal that CoOp favors generalizable and spurious features in the early and later training stages, respectively, leading to the non-overfitting and overfitting phenomena. Given those observations, we propose Subspace Prompt Tuning (Sub PT) to project the gradients in back-propagation onto the low-rank subspace spanned by the early-stage gradient flow eigenvectors during the entire training process and successfully eliminate the overfitting problem. In addition, we equip CoOp with a Novel Feature Learner (NFL) to enhance the generalization ability of the learned prompts onto novel categories beyond the training set, needless of image training data. Extensive experiments on 11 classification datasets demonstrate that Sub PT+NFL consistently boost the performance of CoOp and outperform the state-of-the-art CoCoOp approach. Experiments on more challenging vision downstream tasks, including open-vocabulary object detection and zero-shot semantic segmentation, also verify the effectiveness of the proposed method. Codes can be found athttps://tinyurl.com/mpe64f89.

Salient object detection is a binary pixel-wise classification to distinguish objects in an image, and also have attracted many research interests in the optical Remote Sensing Images (RSIs). The existing state-of-the-art method exploits the full encoder-decoder architecture to predict salient objects in the optical RSIs, suffering from the problem of unsmooth edges and incomplete structures. To address these problems, in this paper, we propose a Boundary-Aware Network (BANet) with two-stage partial decoders sharing the same encoders for salient object detection in RSIs. Specifically, a Boundary-Aware Partial Decoder (BAD) is introduced at the first stage to focus on learning clear edges of salient objects. To solve the pixel-imbalance problem between boundary and background, an edge-aware loss is proposed to guide learning the BAD network. The resulting features are then employed in turn to enhance high-level features. Afterwards, the Structure-Aware Partial Decoder (SAD) is further introduced at the second stage to improve the structure integrity of salient objects. To alleviate the problem of incomplete structures, the structural similarity loss is further proposed to supervise learning the SAD network. In a consequence, our proposed BANet can predict salient objects with clear edges and complete structure, while reducing model parameters due to the discardment of low-level features. Besides, training a deep neural network requires a large amount of images, and the current benchmark datasets for optical remote sensing images are not large enough. Therefore, we also create a large-scale challenging dataset for salient object detection in RSIs. Extensive experiments demonstrate that our proposed BANet outperforms previous RSI SOD models on all existing benchmark datasets and our new presented dataset available at https://github.com/QingpingZheng/RSISOD.

The long-tailed distribution is a common phenomenon in the real world. Extracted large scale image datasets inevitably demonstrate the long-tailed property and models trained with imbalanced data can obtain high performance for the over-represented categories, but struggle for the under-represented categories, leading to biased predictions and performance degradation. To address this challenge, we propose a novel de-biasing method named Inverse Image Frequency (IIF). IIF is a multiplicative margin adjustment transformation of the logits in the classification layer of a convolutional neural network. Our method achieves stronger performance than similar works and it is especially useful for downstream tasks such as long-tailed instance segmentation as it produces fewer false positive detections. Our extensive experiments show that IIF surpasses the state of the art on many long-tailed benchmarks such as ImageNet-LT, CIFAR-LT, Places-LT and LVIS, reaching 55.8% top-1 accuracy with ResNet50 on ImageNet-LT and 26.3% segmentation AP with MaskRCNN ResNet50 on LVIS. Code available at https://github.com/kostas1515/iif.

Perspective distortions and crowd variations make crowd counting a challenging task in computer vision. To tackle it, many previous works have used multi-scale architecture in deep neural networks (DNNs). Multi-scale branches can be either directly merged (e.g. by concatenation) or merged through the guidance of proxies (e.g. attentions) in the DNNs. Despite their prevalence, these combination methods are not sophisticated enough to deal with the per-pixel performance discrepancy over multi-scale density maps. In this work, we redesign the multi-scale neural network by introducing a hierarchical mixture of density experts, which hierarchically merges multi-scale density maps for crowd counting. Within the hierarchical structure, an expert competition and collaboration scheme is presented to encourage contributions from all scales; pixel-wise soft gating nets are introduced to provide pixel-wise soft weights for scale combinations in different hierarchies. The network is optimized using both the crowd density map and the local counting map, where the latter is obtained by local integration on the former. Optimizing both can be problematic because of their potential conflicts. We introduce a new relative local counting loss based on relative count differences among hard-predicted local regions in an image, which proves to be complementary to the conventional absolute error loss on the density map. Experiments show that our method achieves the state-of-the-art performance on five public datasets, i.e. ShanghaiTech, UCF_CC_50, JHU-CROWD++, NWPU-Crowd and Trancos. Our codes will be available at https://github.com/ZPDu/Redesigning-Multi-Scale-Neural-Network-for-Crowd-Counting.

Learning discriminative deep feature embeddings by using million-scale in-the-wild datasets and margin-based softmax loss is the current state-of-the-art approach for face recognition. However, the memory and computing cost of the Fully Connected (FC) layer linearly scales up to the number of identities in the training set. Besides, the largescale training data inevitably suffers from inter-class conflict and long-tailed distribution. In this paper, we propose a sparsely updating variant of the FC layer, named Partial FC (PFC). In each iteration, positive class centers and a random subset of negative class centers are selected to compute the margin-based softmax loss. All class centers are still maintained throughout the whole training process, but only a subset is selected and updated in each iteration. Therefore, the computing requirement, the probability of inter-class conflict, and the frequency of passive update on tail class centers, are dramatically reduced. Extensive experiments across different training data and backbones (e.g. CNN and ViT) confirm the effectiveness, robustness and efficiency of the proposed PFC. The source code is available at https://github.com/deepinsight/insightface/tree/master/recognition.

Benefiting from the pioneering design of generic object detectors, significant achievements have been made in the field of face detection. Typically, the architectures of the backbone, feature pyramid layer, and detection head module within the face detector all assimilate the excellent experience from general object detectors. However, several effective methods, including label assignment and scale-level data augmentation strategy, fail to maintain consistent superiority when applying on the face detector directly. Concretely, the former strategy involves a vast body of hyperparameters and the latter one suffers from the challenge of scale distribution bias between different detection tasks, which both limit their generalization abilities. Furthermore, in order to provide accurate face bounding boxes for facial down-stream tasks, the face detector imperatively requires the elimination of false alarms. As a result, practical solutions on label assignment, scale-level data augmentation, and reducing false alarms are necessary for advancing face detectors. In this paper, we focus on resolving three aforementioned challenges that exiting methods are difficult to finish off and present a novel face detector, termed MogFace. In our Mogface, three key components, Adaptive Online Incremental Anchor Mining Strategy, Selective Scale Enhancement Strategy and Hierarchical Context-Aware Module, are separately proposed to boost the performance of face detectors. Finally, to the best of our knowledge, our MogFace is the best face detector on the Wider Face leader-board, achieving all champions across different testing scenarios. The code is available at https://github.com/damo-cv/MogFace.

This paper probes intrinsic factors behind typical failure cases (e.g. spatial inconsistency and boundary confusion) produced by the existing state-of-the-art method in face parsing. To tackle these problems, we propose a novel Decoupled Multi-task Learning with Cyclical Self-Regulation (DML-CSR) for face parsing. Specifically, DML-CSR designs a multi-task model which comprises face parsing, binary edge, and category edge detection. These tasks only share low-level encoder weights without high-level interactions between each other, enabling to decouple auxiliary modules from the whole network at the inference stage. To address spatial inconsistency, we develop a dynamic dual graph convolutional network to capture global contextual information without using any extra pooling operation. To handle boundary confusion in both single and multiple face scenarios, we exploit binary and category edge detection to jointly obtain generic geometric structure and fine-grained semantic clues of human faces. Besides, to prevent noisy labels from degrading model generalization during training, cyclical self-regulation is proposed to self-ensemble several model instances to get a new model and the resulting model then is used to self-distill subsequent models, through alternating iterations. Experiments show that our method achieves the new state-of-the-art performance on the Helen, CelebAMask-HQ, and Lapa datasets. The source code is available at https://github.com/deepinsight/insightface/tree/master/parsing/dml_csr.

Near infrared (NIR) to Visible (VIS) face matching is challenging due to the significant domain gaps as well as a lack of sufficient data for cross-modality model training. To overcome this problem, we propose a novel method for paired NIR-VIS facial image generation. Specifically, we reconstruct 3D face shape and reflectance from a large 2D facial dataset and introduce a novel method of transforming the VIS reflectance to NIR reflectance. We then use a physically-based renderer to generate a vast, high-resolution and photorealistic dataset consisting of various poses and identities in the NIR and VIS spectra. Moreover, to facilitate the identity feature learning, we propose an IDentity-based Maximum Mean Discrepancy (ID-MMD) loss, which not only reduces the modality gap between NIR and VIS images at the domain level but encourages the network to focus on the identity features instead of facial details, such as poses and accessories. Extensive experiments conducted on four challenging NIR-VIS face recognition benchmarks demonstrate that the proposed method can achieve comparable performance with the state-of-the-art (SOTA) methods without requiring any existing NIR-VIS face recognition datasets. With slightly fine-tuning on the target NIR-VIS face recognition datasets, our method can significantly surpass the SOTA performance. Code and pretrained models are released under the insightface GitHub.

Deep convolutional neural networks (DCNNs) are currently the method of choice both for generative, as well as for discriminative learning in computer vision and machine learning. The success of DCNNs can be attributed to the careful selection of their building blocks (e.g., residual blocks, rectifiers, sophisticated normalization schemes, to mention but a few). In this paper, we propose Π-Nets, a new class of function approximators based on polynomial expansions. Π-Nets are polynomial neural networks, i.e., the output is a high-order polynomial of the input. The unknown parameters, which are naturally represented by high-order tensors, are estimated through a collective tensor factorization with factors sharing. We introduce three tensor decompositions that significantly reduce the number of parameters and show how they can be efficiently implemented by hierarchical neural networks. We empirically demonstrate that Π-Nets are very expressive and they even produce good results without the use of non-linear activation functions in a large battery of tasks and signals, i.e., images, graphs, and audio. When used in conjunction with activation functions, Π-Nets produce state-of-the-art results in three challenging tasks, i.e., image generation, face verification and 3D mesh representation learning. The source code is available at https://github.com/grigorisg9gr/polynomial_nets.

Deep face recognition has achieved remarkable improvements due to the introduction of margin-based softmax loss, in which the prototype stored in the last linear layer represents the center of each class. In these methods, training samples are enforced to be close to positive prototypes and far apart from negative prototypes by a clear margin. However, we argue that prototype learning only employs sample-to-prototype comparisons without considering sample-to-sample comparisons during training and the low loss value gives us an illusion of perfect feature embedding, impeding the further exploration of SGD. To this end, we propose Variational Prototype Learning (VPL), which represents every class as a distribution instead of a point in the latent space. By identifying the slow feature drift phenomenon, we directly inject memorized features into prototypes to approximate variational prototype sampling. The proposed VPL can simulate sample-to-sample comparisons within the classification framework, encouraging the SGD solver to be more exploratory, while boosting performance. Moreover, VPL is conceptually simple, easy to implement, computationally efficient and memory saving. We present extensive experimental results on popular benchmarks, which demonstrate the superiority of the proposed VPL method over the state-of-the-art competitors.

The last few years have witnessed the great success of non-linear generative models in synthesizing high-quality photorealistic face images. Many recent 3D facial texture reconstruction and pose manipulation from a single image approaches still rely on large and clean face datasets to train image-to-image Generative Adversarial Networks (GANs). Yet the collection of such a large scale high-resolution 3D texture dataset is still very costly and difficult to maintain age/ethnicity balance. Moreover, regression-based approaches suffer from generalization to the in-the-wild conditions and are unable to fine-tune to a target-image. In this work, we propose an unsupervised approach for one-shot 3D facial texture completion that does not re-quire large-scale texture datasets, but rather harnesses the knowledge stored in 2D face generators. The proposed approach rotates an input image in 3D and fill-in the unseen regions by reconstructing the rotated image in a 2D face generator, based on the visible parts. Finally, we stitch the most visible textures at different angles in the UV image-plane. Further, we frontalize the target image by projecting the completed texture into the generator. The qualitative and quantitative experiments demonstrate that the completed UV textures and frontalized images are of high quality, resembles the original identity, can be used to train a texture GAN model for 3DMM fitting and improve pose-invariant face recognition.1

In this paper, we contribute a new million-scale face benchmark containing noisy 4M identities/260M faces (WebFace260M) and cleaned 2M identities/42M faces (WebFace42M) training data, as well as an elaborately designed time-constrained evaluation protocol. Firstly, we collect 4M name list and download 260M faces from the Internet. Then, a Cleaning Automatically utilizing Self-Training (CAST) pipeline is devised to purify the tremendous WebFace260M, which is efficient and scalable. To the best of our knowledge, the cleaned WebFace42M is the largest public face recognition training set and we expect to close the data gap between academia and industry. Referring to practical scenarios, Face Recognition Under Inference Time conStraint (FRUITS) protocol and a test set are constructed to comprehensively evaluate face matchers.Equipped with this benchmark, we delve into million-scale face recognition problems. A distributed framework is developed to train face recognition models efficiently without tampering with the performance. Empowered by Web-Face42M, we reduce relative 40% failure rate on the challenging IJB-C set, and rank the 3rd among 430 entries on NIST-FRVT. Even 10% data (WebFace4M) shows superior performance compared with public training set. Furthermore, comprehensive baselines are established on our rich-attribute test set under FRUITS-100ms/500ms/1000ms protocol, including MobileNet, EfficientNet, AttentionNet, ResNet, SENet, ResNeXt and RegNet families. Benchmark website is https://www.face-benchmark.org.

Spatial self-attention layers, in the form of Non-Local blocks, introduce long-range dependencies in Convolutional Neural Networks by computing pairwise similarities among all possible positions. Such pairwise functions underpin the effectiveness of non-local layers, but also determine a complexity that scales quadratically with respect to the input size both in space and time. This is a severely limiting factor that practically hinders the applicability of non-local blocks to even moderately sized inputs. Previous works focused on reducing the complexity by modifying the underlying matrix operations, however in this work we aim to retain full expressiveness of non-local layers while keeping complexity linear. We overcome the efficiency limitation of non-local blocks by framing them as special cases of 3rd order polynomial functions. This fact enables us to formulate novel fast Non-Local blocks, capable of reducing the complexity from quadratic to linear with no loss in performance, by replacing any direct computation of pairwise similarities with element-wise multiplications. The proposed method, which we dub as "Poly-NL", is competitive with state-of-the-art performance across image recognition, instance segmentation, and face detection tasks, while having considerably less computational overhead.

Deep Convolutional Neural Networks (DCNNs) is currently the method of choice both for generative, as well as for discriminative learning in computer vision and machine learning. The success of DCNNs can be attributed to the careful selection of their building blocks (e.g., residual blocks, rectifiers, sophisticated normalization schemes, to mention but a few). In this paper, we propose Π-Nets, a new class of DCNNs. Π-Nets are polynomial neural networks, i.e., the output is a high-order polynomial of the input. Π-Nets can be implemented using special kind of skip connections and their parameters can be represented via high-order tensors. We empirically demonstrate that Π-Nets have better representation power than standard DCNNs and they even produce good results without the use of non-linear activation functions in a large battery of tasks and signals, i.e., images, graphs, and audio. When used in conjunction with activation functions, Π-Nets produce state-of-the-art results in challenging tasks, such as image generation. Lastly, our framework elucidates why recent generative models, such as StyleGAN, improve upon their predecessors, e.g., ProGAN.

Though tremendous strides have been made in uncontrolled face detection, accurate and efficient 2D face alignment and 3D face reconstruction in-the-wild remain an open challenge. In this paper, we present a novel single-shot, multi-level face localisation method, named RetinaFace, which unifies face box prediction, 2D facial landmark localisation and 3D vertices regression under one common target: point regression on the image plane. To fill the data gap, we manually annotated five facial landmarks on the WIDER FACE dataset and employed a semi-automatic annotation pipeline to generate 3D vertices for face images from the WIDER FACE, AFLW and FDDB datasets. Based on extra annotations, we propose a mutually beneficial regression target for 3D face reconstruction, that is predicting 3D vertices projected on the image plane constrained by a common 3D topology. The proposed 3D face reconstruction branch can be easily incorporated, without any optimisation difficulty, in parallel with the existing box and 2D landmark regression branches during joint training. Extensive experimental results show that RetinaFace can simultaneously achieve stable face detection, accurate 2D face alignment and robust 3D face reconstruction while being efficient through single-shot inference.

Neural architecture search (NAS) advances beyond the state-of-the-art in various computer vision tasks by automating the designs of deep neural networks. In this paper, we aim to address three important questions in NAS: (1) How to measure the correlation between architectures and their performances? (2) How to evaluate the correlation between different architectures? (3) How to learn these correlations with a small number of samples? To this end, we first model these correlations from a Bayesian perspective. Specifically, by introducing a novel Gaussian Process based NAS (GP-NAS) method, the correlations are modeled by the kernel function and mean function. The kernel function is also learnable to enable adaptive modeling for complex correlations in different search spaces. Furthermore, by incorporating a mutual information based sampling method, we can theoretically ensure the high-performance architecture with only a small set of samples. After addressing these problems, training GP-NAS once enables direct performance prediction of any architecture in different scenarios and may obtain efficient networks for different deployment platforms. Extensive experiments on both image classification and face recognition tasks verify the effectiveness of our algorithm.

The transition matrix, denoting the transition relationship from clean labels to noisy labels, is essential to build statistically consistent classifiers in label-noise learning. Existing methods for estimating the transition matrix rely heavily on estimating the noisy class posterior. However, the estimation error for noisy class posterior could be large because of the randomness of label noise. The estimation error would lead the transition matrix to be poorly estimated. Therefore in this paper, we aim to solve this problem by exploiting the divide-and-conquer paradigm. Specifically, we introduce an intermediate class to avoid directly estimating the noisy class posterior. By this intermediate class, the original transition matrix can then be factorized into the product of two easy-to-estimated transition matrices. We term the proposed method as the dual $T$-estimator. Both theoretical analyses and empirical results illustrate the effectiveness of the dual $T$-estimator for estimating transition matrices, leading to better classification performances.

Recently, a popular line of research in face recognition is adopting margins in the well-established softmax loss function to maximize class separability. In this paper, we first introduce an Additive Angular Margin Loss (ArcFace), which not only has a clear geometric interpretation but also significantly enhances the discriminative power. Since ArcFace is susceptible to the massive label noise, we further propose sub-center ArcFace, in which each class contains K sub-centers and training samples only need to be close to any of the K positive sub-centers. Sub-center ArcFace encourages one dominant sub-class that contains the majority of clean faces and non-dominant sub-classes that include hard or noisy faces. Based on this self-propelled isolation, we boost the performance through automatically purifying raw web faces under massive real-world noise. Besides discriminative feature embedding, we also explore the inverse problem, mapping feature vectors to face images. Without training any additional generator or discriminator, the pre-trained ArcFace model can generate identity-preserved face images for both subjects inside and outside the training data only by using the network gradient and Batch Normalization (BN) priors. Extensive experiments demonstrate that ArcFace can enhance the discriminative feature embedding as well as strengthen the generative face synthesis.

3D Morphable Models (3DMMs) are statistical models that represent facial texture and shape variations using a set of linear bases and more particular Principal Component Analysis (PCA). 3DMMs were used as statistical priors for reconstructing 3D faces from images by solving non-linear least square optimization problems. Recently, 3DMMs were used as generative models for training non-linear mappings (\ie, regressors) from image to the parameters of the models via Deep Convolutional Neural Networks (DCNNs). Nevertheless, all of the above methods use either fully connected layers or 2D convolutions on parametric unwrapped UV spaces leading to large networks with many parameters. In this paper, we present the first, to the best of our knowledge, non-linear 3DMMs by learning joint texture and shape auto-encoders using direct mesh convolutions. We demonstrate how these auto-encoders can be used to train very light-weight models that perform Coloured Mesh Decoding (CMD) in-the-wild at a speed of over 2500 FPS.

In this article, we present the Menpo 2D and Menpo 3D benchmarks, two new datasets for multi-pose 2D and 3D facial landmark localisation and tracking. In contrast to the previous benchmarks such as 300W and 300VW, the proposed benchmarks contain facial images in both semi-frontal and profile pose. We introduce an elaborate semi-automatic methodology for providing high-quality annotations for both the Menpo 2D and Menpo 3D benchmarks. In Menpo 2D benchmark, different visible landmark configurations are designed for semi-frontal and profile faces, thus making the 2D face alignment full-pose. In Menpo 3D benchmark, a united landmark configuration is designed for both semi-frontal and profile faces based on the correspondence with a 3D face model, thus making face alignment not only full-pose but also corresponding to the real-world 3D space. Based on the considerable number of annotated images, we organised Menpo 2D Challenge and Menpo 3D Challenge for face alignment under large pose variations in conjunction with CVPR 2017 and ICCV 2017, respectively. The results of these challenges demonstrate that recent deep learning architectures, when trained with the abundant data, lead to excellent results. We also provide a very simple, yet effective solution, named Cascade Multi-view Hourglass Model, to 2D and 3D face alignment. In our method, we take advantage of all 2D and 3D facial landmark annotations in a joint way. We not only capitalise on the correspondences between the semi-frontal and profile 2D facial landmarks but also employ joint supervision from both 2D and 3D facial landmarks. Finally, we discuss future directions on the topic of face alignment.

Robust principal component analysis (RPCA) is a powerful method for learning low-rank feature representation of various visual data. However, for certain types as well as significant amount of error corruption, it fails to yield satisfactory results; a drawback that can be alleviated by exploiting domain-dependent prior knowledge or information. In this paper, we propose two models for the RPCA that take into account such side information, even in the presence of missing values. We apply this framework to the task of UV completion which is widely used in pose-invariant face recognition. Moreover, we construct a generative adversarial network (GAN) to extract side information as well as subspaces. These subspaces not only assist in the recovery but also speed up the process in case of large-scale data. We quantitatively and qualitatively evaluate the proposed approaches through both synthetic data and eight real-world datasets to verify their effectiveness.

The de facto algorithm for facial landmark estimation involves running a face detector with a subsequent deformable model fitting on the bounding box. This encompasses two basic problems: i) the detection and deformable fitting steps are performed independently, while the detector might not provide best-suited initialization for the fitting step, ii) the face appearance varies hugely across different poses, which makes the deformable face fitting very challenging and thus distinct models have to be used (e.g., one for profile and one for frontal faces). In this work, we propose the first, to the best of our knowledge, joint multi-view convolutional network to handle large pose variations across faces in-the-wild, and elegantly bridge face detection and facial landmark localization tasks. Existing joint face detection and landmark localization methods focus only on a very small set of landmarks. By contrast, our method can detect and align a large number of landmarks for semi-frontal (68 landmarks) and profile (39 landmarks) faces. We evaluate our model on a plethora of datasets including standard static image datasets such as IBUG, 300W, COFW, and the latest Menpo Benchmark for both semi-frontal and profile faces. Significant improvement over state-of-the-art methods on deformable face tracking is witnessed on 300VW benchmark. We also demonstrate state-ofthe- art results for face detection on FDDB and MALF datasets.

We revisit the problem of robust principal component analysis with features acting as prior side information. To this aim, a novel, elegant, non-convex optimization approach is proposed to decompose a given observation matrix into a low-rank core and the corresponding sparse residual. Rigorous theoretical analysis of the proposed algorithm results in exact recovery guarantees with low computational complexity. Aptly designed synthetic experiments demonstrate that our method is the first to wholly harness the power of non-convexity over convexity in terms of both recoverability and speed. That is, the proposed non-convex approach is more accurate and faster compared to the best available algorithms for the problem under study. Two real-world applications, namely image classification and face denoising further exemplify the practical superiority of the proposed method.

Recently proposed robust 3D face alignment methods establish either dense or sparse correspondence between a 3D face model and a 2D facial image. The use of these methods presents new challenges as well as opportunities for facial texture analysis. In particular, by sampling the image using the fitted model, a facial UV can be created. Unfortunately, due to self-occlusion, such a UV map is always incomplete. In this paper, we propose a framework for training Deep Convolutional Neural Network (DCNN) to complete the facial UV map extracted from in-the-wild images. To this end, we first gather complete UV maps by fitting a 3D Morphable Model (3DMM) to various multiview image and video datasets, as well as leveraging on a new 3D dataset with over 3,000 identities. Second, we devise a meticulously designed architecture that combines local and global adversarial DCNNs to learn an identity-preserving facial UV completion model. We demonstrate that by attaching the completed UV to the fitted mesh and generating instances of arbitrary poses, we can increase pose variations for training deep face recognition/verification models, and minimise pose discrepancy during testing, which lead to better performance. Experiments on both controlled and in-the-wild UV datasets prove the effectiveness of our adversarial UV completion model. We achieve state-of-the-art verification accuracy, 94.05%, under the CFP frontal-profile protocol only by combining pose augmentation during training and pose discrepancy reduction during testing. We will release the first in-the-wild UV dataset (we refer as WildUV) that comprises of complete facial UV maps from 1,892 identities for research purposes.

Estimating the 3D facial landmarks from a 2D image remains a challenging problem. Even though state-of-the-art 2D alignment methods are able to predict accurate landmarks for semi-frontal faces, the majority of them fail to provide semantically consistent landmarks for profile faces. A de facto solution to this problem is through 3D face alignment that preserves correspondence across different poses. In this paper, we proposed a Cascade Multi-view Hourglass Model for 3D face alignment, where the first Hourglass model is explored to jointly predict semi-frontal and profile 2D facial landmarks, after removing spatial transformations, another Hourglass model is employed to estimate the 3D facial shapes. To improve the capacity without sacrificing the computational complexity, the original residual bottleneck block in the Hourglass model is replaced by a parallel, multi-scale inception-resnet block. Extensive experiments on two challenging 3D face alignment datasets, AFLW2000-3D and Menpo-3D, show the robustness of the proposed method under continuous pose changes.

In this work we explore the use of shape-based representations as an auxiliary source of supervision for pose estimation and action recognition. We show that shape-based representations can act as a source of `privileged information' that complements and extends the pure landmark-level annotations. We explore 2D shape-based supervision signals, such as Support Vector Shape. Our experiments show that shape-based supervision signals substantially improve pose alignment accuracy in the form of a cascade architecture. We outperform state-of-the-art methods on the MPII and LSP datasets, while using substantially shallower networks. For action localization in untrimmed videos, our method introduces additional classification signals based on the structured segment networks (SSN) and further improved the performance. To be specific, dense human pose and landmarks localization signals are involved in detection progress. We applied out network to all frames of videos alongside with output from SSN to further improve detection accuracy, especially for pose related and sparsely annotated videos. The method in general achieves state-of-the-art performance on Activity Detection Task for ActivityNet Challenge2017 test set and witnesses remarkable improvement on pose related and sparsely annotated categories e.g. sports.

Cascade regression is a popular face alignment approach, and it has achieved good performances on the wild databases. However, it depends heavily on local features in estimating reliable landmark locations and therefore suffers from corrupted images, such as images with occlusion, which often exists in real-world face images. In this paper, we present a new adaptive cascade regression model for robust face alignment. In each iteration, the shape-indexed appearance is introduced to estimate the occlusion level of each landmark, and each landmark is then weighted according to its estimated occlusion level. Also, the occlusion levels of the landmarks act as adaptive weights on the shape-indexed features to decrease the noise on the shape-indexed features. At the same time, an exemplar-based shape prior is designed to suppress the influence of local image corruption. Extensive experiments are conducted on the challenging benchmarks, and the experimental results demonstrate that the proposed method achieves better results than the state-of-the-art methods for facial landmark localization and occlusion detection.

Localizing facial landmarks is a fundamental step in facial image analysis. However, the problem continues to be challenging due to the large variability in expression, illumination, pose, and the existence of occlusions in real-world face images. In this paper, we present a dual sparse constrained cascade regression model for robust face alignment. Instead of using the least-squares method during the training process of regressors, sparse constraint is introduced to select robust features and compress the size of the model. Moreover, sparse shape constraint is incorporated between each cascade regression, and the explicit shape constraints are able to suppress the ambiguity in local features. To improve the model's adaptation to large pose variation, face pose is estimated by five fiducial landmarks located by deep convolutional neuron network, which is used to adaptively design the cascade regression model. To the best of our best knowledge, this is the first attempt to fuse explicit shape constraint (sparse shape constraint) and implicit context information (sparse feature selection) for robust face alignment in the framework of cascade regression. Extensive experiments on nine challenging wild data sets demonstrate the advantages of the proposed method over the state-of-the-art methods.

Editing faces in videos is a popular yet challenging task in computer vision and graphics that encompasses various applications, including facial attractiveness enhancement, makeup transfer, face replacement, and expression manipulation. Directly applying the existing warping methods to video face editing has the major problem of temporal incoherence in the synthesized videos, which cannot be addressed by simply employing face tracking techniques or manual interventions, as it is difficult to eliminate the subtly temporal incoherence of the facial feature point localizations in a video sequence. In this article, we propose a temporal-spatial-smooth warping (TSSW) method to achieve a high temporal coherence for video face editing. TSSW is based on two observations: (1) the control lattices are critical for generating warping surfaces and achieving the temporal coherence between consecutive video frames, and (2) the temporal coherence and spatial smoothness of the control lattices can be simultaneously and effectively preserved. Based upon these observations, we impose the temporal coherence constraint on the control lattices on two consecutive frames, as well as the spatial smoothness constraint on the control lattice on the current frame. TSSW calculates the control lattice (in either the horizontal or vertical direction) by updating the control lattice (in the corresponding direction) on its preceding frame, i.e., minimizing a novel energy function that unifies a data-driven term, a smoothness term, and feature point constraints. The contributions of this article are twofold: (1) we develop TSSW, which is robust to the subtly temporal incoherence of the facial feature point localizations and is effective to preserve the temporal coherence and spatial smoothness of the control lattices for editing faces in videos, and (2) we present a new unified video face editing framework that is capable for improving the performances of facial attractiveness enhancement, makeup transfer, face replacement, and expression manipulation.

Explicit feature mapping is an appealing way to linearize additive kernels, such as χ2kernel for training large-scale support vector machines (SVMs). Although accurate in approximation, feature mapping could pose computational challenges in high-dimensional settings as it expands the original features to a higher dimensional space. To handle this issue in the context of χ2kernel SVMs learning, we introduce a simple yet efficient method to approximately linearize χ2kernel through random feature maps. The main idea is to use sparse random projection to reduce the dimensionality of feature maps while preserving their approximation capability to the original kernel. We provide approximation error bound for the proposed method. Furthermore, we extend our method to χ2multiple kernel SVMs learning. Extensive experiments on large-scale image classification tasks confirm that the proposed approach is able to significantly speed up the training process of the χ2kernel SVMs at almost no cost of testing accuracy.