Hongdong Li

Junlin Han, Lars Petersson, Hongdong Li et al. · oxford

We present a simple method, CropMix, for the purpose of producing a rich input distribution from the original dataset distribution. Unlike single random cropping, which may inadvertently capture only limited information, or irrelevant information, like pure background, unrelated objects, etc, we crop an image multiple times using distinct crop scales, thereby ensuring that multi-scale information is captured. The new input distribution, serving as training data, useful for a number of vision tasks, is then formed by simply mixing multiple cropped views. We first demonstrate that CropMix can be seamlessly applied to virtually any training recipe and neural network architecture performing classification tasks. CropMix is shown to improve the performance of image classifiers on several benchmark tasks across-the-board without sacrificing computational simplicity and efficiency. Moreover, we show that CropMix is of benefit to both contrastive learning and masked image modeling towards more powerful representations, where preferable results are achieved when learned representations are transferred to downstream tasks. Code is available at GitHub.

Ruyi Zha, Xuelian Cheng, Hongdong Li et al. · ibm-research

Reconstructing soft tissues from stereo endoscope videos is an essential prerequisite for many medical applications. Previous methods struggle to produce high-quality geometry and appearance due to their inadequate representations of 3D scenes. To address this issue, we propose a novel neural-field-based method, called EndoSurf, which effectively learns to represent a deforming surface from an RGBD sequence. In EndoSurf, we model surface dynamics, shape, and texture with three neural fields. First, 3D points are transformed from the observed space to the canonical space using the deformation field. The signed distance function (SDF) field and radiance field then predict their SDFs and colors, respectively, with which RGBD images can be synthesized via differentiable volume rendering. We constrain the learned shape by tailoring multiple regularization strategies and disentangling geometry and appearance. Experiments on public endoscope datasets demonstrate that EndoSurf significantly outperforms existing solutions, particularly in reconstructing high-fidelity shapes. Code is available at https://github.com/Ruyi-Zha/endosurf.git.

Yuwei Qiu, Kaihao Zhang, Chenxi Wang et al.

In recent years, Transformer networks are beginning to replace pure convolutional neural networks (CNNs) in the field of computer vision due to their global receptive field and adaptability to input. However, the quadratic computational complexity of softmax-attention limits the wide application in image dehazing task, especially for high-resolution images. To address this issue, we propose a new Transformer variant, which applies the Taylor expansion to approximate the softmax-attention and achieves linear computational complexity. A multi-scale attention refinement module is proposed as a complement to correct the error of the Taylor expansion. Furthermore, we introduce a multi-branch architecture with multi-scale patch embedding to the proposed Transformer, which embeds features by overlapping deformable convolution of different scales. The design of multi-scale patch embedding is based on three key ideas: 1) various sizes of the receptive field; 2) multi-level semantic information; 3) flexible shapes of the receptive field. Our model, named Multi-branch Transformer expanded by Taylor formula (MB-TaylorFormer), can embed coarse to fine features more flexibly at the patch embedding stage and capture long-distance pixel interactions with limited computational cost. Experimental results on several dehazing benchmarks show that MB-TaylorFormer achieves state-of-the-art (SOTA) performance with a light computational burden. The source code and pre-trained models are available at https://github.com/FVL2020/ICCV-2023-MB-TaylorFormer.

Siva Rajesh Kasa, Yasong Dai, Sumit Negi et al.

Diffusion large language models promise parallel token generation, yet inference remains bottlenecked by deciding which masked tokens can be safely committed together. Fast-dLLM addressed this with KV caching and confidence-guided parallel decoding, but its decoding theory uses a homogeneous high-confidence assumption that effectively reduces each candidate set to its weakest selected token. We argue that this leaves speed on the table because real decoding steps exhibit heterogeneous confidence profiles. We propose \textbf{Fast-dLLM++}, a training-free extension that introduces \emph{Fréchet profile decoding}: selecting parallel commit sets from the full sorted confidence profile rather than a single worst-case confidence. The resulting rule is a heterogeneous-confidence generalization of Fast-dLLM's factor selector and it recovers the previous rule exactly in the equal-confidence case and adds a provable \emph{heterogeneity bonus} when the selected tokens have uneven confidences. Fast-dLLM++ leaves the model, diffusion process, and cache implementation entirely unchanged, making it a drop-in replacement for existing Fast-dLLM decoding. Experiments on GSM8K, MATH, HumanEval, and MBPP with the LLaDA-8B model show that the theoretical improvement translates directly into empirical gains: profile-aware selection improves the accuracy--throughput frontier by exploiting safe parallelism that weakest-token rules miss, achieving up to 37\% higher throughput at comparable accuracy. Our anonymous code release is at https://github.com/Ringo-Star/FastdLLM_plusplus.

Rong Wang, Wei Mao, Hongdong Li

3D hand-object pose estimation is the key to the success of many computer vision applications. The main focus of this task is to effectively model the interaction between the hand and an object. To this end, existing works either rely on interaction constraints in a computationally-expensive iterative optimization, or consider only a sparse correlation between sampled hand and object keypoints. In contrast, we propose a novel dense mutual attention mechanism that is able to model fine-grained dependencies between the hand and the object. Specifically, we first construct the hand and object graphs according to their mesh structures. For each hand node, we aggregate features from every object node by the learned attention and vice versa for each object node. Thanks to such dense mutual attention, our method is able to produce physically plausible poses with high quality and real-time inference speed. Extensive quantitative and qualitative experiments on large benchmark datasets show that our method outperforms state-of-the-art methods. The code is available at https://github.com/rongakowang/DenseMutualAttention.git.

Jiaxin Wei, Xibin Song, Weizhe Liu et al.

While showing promising results, recent RGB-D camera-based category-level object pose estimation methods have restricted applications due to the heavy reliance on depth sensors. RGB-only methods provide an alternative to this problem yet suffer from inherent scale ambiguity stemming from monocular observations. In this paper, we propose a novel pipeline that decouples the 6D pose and size estimation to mitigate the influence of imperfect scales on rigid transformations. Specifically, we leverage a pre-trained monocular estimator to extract local geometric information, mainly facilitating the search for inlier 2D-3D correspondence. Meanwhile, a separate branch is designed to directly recover the metric scale of the object based on category-level statistics. Finally, we advocate using the RANSAC-P$n$P algorithm to robustly solve for 6D object pose. Extensive experiments have been conducted on both synthetic and real datasets, demonstrating the superior performance of our method over previous state-of-the-art RGB-based approaches, especially in terms of rotation accuracy. Code: https://github.com/goldoak/DMSR.

Tao Wang, Kaihao Zhang, Ziqian Shao et al.

Current deep learning methods for low-light image enhancement (LLIE) typically rely on pixel-wise mapping learned from paired data. However, these methods often overlook the importance of considering degradation representations, which can lead to sub-optimal outcomes. In this paper, we address this limitation by proposing a degradation-aware learning scheme for LLIE using diffusion models, which effectively integrates degradation and image priors into the diffusion process, resulting in improved image enhancement. Our proposed degradation-aware learning scheme is based on the understanding that degradation representations play a crucial role in accurately modeling and capturing the specific degradation patterns present in low-light images. To this end, First, a joint learning framework for both image generation and image enhancement is presented to learn the degradation representations. Second, to leverage the learned degradation representations, we develop a Low-Light Diffusion model (LLDiffusion) with a well-designed dynamic diffusion module. This module takes into account both the color map and the latent degradation representations to guide the diffusion process. By incorporating these conditioning factors, the proposed LLDiffusion can effectively enhance low-light images, considering both the inherent degradation patterns and the desired color fidelity. Finally, we evaluate our proposed method on several well-known benchmark datasets, including synthetic and real-world unpaired datasets. Extensive experiments on public benchmarks demonstrate that our LLDiffusion outperforms state-of-the-art LLIE methods both quantitatively and qualitatively. The source code and pre-trained models are available at https://github.com/TaoWangzj/LLDiffusion.

Hanlin Chen, Jiaxin Wei, Xibin Song et al.

Frame-wise action-controlled image-to-video generation is a promising paradigm for interactive world simulation, where each control signal should elicit an immediate visual response. However, maintaining visual fidelity and 3D consistency over long autoregressive rollouts remains challenging. Existing 3D-aware methods often suffer from catastrophic drift due to two impediments: information loss from \textit{Latent--RGB Cycling}, where generated latents are repeatedly decoded to RGB and re-encoded for future conditioning, and the training--inference gap induced by the \textit{error-free hypothesis}, where clean training memory fails to match prediction-corrupted inference memory. To address these challenges, we present \textbf{Robust Dreamer}, a memory-augmented framework built around how to design 3D memory and how to use it robustly. First, we introduce \textbf{Latent Gaussian Memory}, which anchors diffusion latents inherited from the generation process to Gaussian primitives and recalls them via latent-space Gaussian splatting. This provides dense, geometry-aware, view-aligned conditioning while avoiding accumulated degradation from repeated VAE conversion. Second, we propose \textbf{Deviation Learning with Dynamic Deviation Archive}, which synthesizes rollout-induced latent deviations through a one-step approximation, stores them by autoregressive stage and denoising timestamp, and injects them into historical memory during training. This exposes the generator to realistic corrupted memory states and teaches internal correction before inference. Experiments on ScanNet, DL3DV, and OmniWorldGame demonstrate state-of-the-art long-horizon performance.

Huan Lei, Ruitao Leng, Liang Zheng et al.

Reconstructing 3D point clouds into triangle meshes is a key problem in computational geometry and surface reconstruction. Point cloud triangulation solves this problem by providing edge information to the input points. Since no vertex interpolation is involved, it is beneficial to preserve sharp details on the surface. Taking advantage of learning-based techniques in triangulation, existing methods enumerate the complete combinations of candidate triangles, which is both complex and inefficient. In this paper, we leverage the duality between a triangle and its circumcenter, and introduce a deep neural network that detects the circumcenters to achieve point cloud triangulation. Specifically, we introduce multiple anchor priors to divide the neighborhood space of each point. The neural network then learns to predict the presences and locations of circumcenters under the guidance of those anchors. We extract the triangles dual to the detected circumcenters to form a primitive mesh, from which an edge-manifold mesh is produced via simple post-processing. Unlike existing learning-based triangulation methods, the proposed method bypasses an exhaustive enumeration of triangle combinations and local surface parameterization. We validate the efficiency, generalization, and robustness of our method on prominent datasets of both watertight and open surfaces. The code and trained models are provided at https://github.com/EnyaHermite/CircNet.

Jiayu Yang, Ziang Cheng, Yunfei Duan et al.

Given a single image of a 3D object, this paper proposes a novel method (named ConsistNet) that is able to generate multiple images of the same object, as if seen they are captured from different viewpoints, while the 3D (multi-view) consistencies among those multiple generated images are effectively exploited. Central to our method is a multi-view consistency block which enables information exchange across multiple single-view diffusion processes based on the underlying multi-view geometry principles. ConsistNet is an extension to the standard latent diffusion model, and consists of two sub-modules: (a) a view aggregation module that unprojects multi-view features into global 3D volumes and infer consistency, and (b) a ray aggregation module that samples and aggregate 3D consistent features back to each view to enforce consistency. Our approach departs from previous methods in multi-view image generation, in that it can be easily dropped-in pre-trained LDMs without requiring explicit pixel correspondences or depth prediction. Experiments show that our method effectively learns 3D consistency over a frozen Zero123 backbone and can generate 16 surrounding views of the object within 40 seconds on a single A100 GPU. Our code will be made available on https://github.com/JiayuYANG/ConsistNet

Yujiao Shi, Fei Wu, Akhil Perincherry et al.

Image retrieval-based cross-view localization methods often lead to very coarse camera pose estimation, due to the limited sampling density of the database satellite images. In this paper, we propose a method to increase the accuracy of a ground camera's location and orientation by estimating the relative rotation and translation between the ground-level image and its matched/retrieved satellite image. Our approach designs a geometry-guided cross-view transformer that combines the benefits of conventional geometry and learnable cross-view transformers to map the ground-view observations to an overhead view. Given the synthesized overhead view and observed satellite feature maps, we construct a neural pose optimizer with strong global information embedding ability to estimate the relative rotation between them. After aligning their rotations, we develop an uncertainty-guided spatial correlation to generate a probability map of the vehicle locations, from which the relative translation can be determined. Experimental results demonstrate that our method significantly outperforms the state-of-the-art. Notably, the likelihood of restricting the vehicle lateral pose to be within 1m of its Ground Truth (GT) value on the cross-view KITTI dataset has been improved from $35.54\%$ to $76.44\%$, and the likelihood of restricting the vehicle orientation to be within $1^{\circ}$ of its GT value has been improved from $19.64\%$ to $99.10\%$.

Junlin Han, Huangying Zhan, Jie Hong et al. · oxford

This paper studies the problem of measuring and predicting how memorable an image is to pattern recognition machines, as a path to explore machine intelligence. Firstly, we propose a self-supervised machine memory quantification pipeline, dubbed ``MachineMem measurer'', to collect machine memorability scores of images. Similar to humans, machines also tend to memorize certain kinds of images, whereas the types of images that machines and humans memorize are different. Through in-depth analysis and comprehensive visualizations, we gradually unveil that``complex" images are usually more memorable to machines. We further conduct extensive experiments across 11 different machines (from linear classifiers to modern ViTs) and 9 pre-training methods to analyze and understand machine memory. This work proposes the concept of machine memorability and opens a new research direction at the interface between machine memory and visual data.

Ruyi Zha, Yanhao Zhang, Hongdong Li

This paper proposes a novel and fast self-supervised solution for sparse-view CBCT reconstruction (Cone Beam Computed Tomography) that requires no external training data. Specifically, the desired attenuation coefficients are represented as a continuous function of 3D spatial coordinates, parameterized by a fully-connected deep neural network. We synthesize projections discretely and train the network by minimizing the error between real and synthesized projections. A learning-based encoder entailing hash coding is adopted to help the network capture high-frequency details. This encoder outperforms the commonly used frequency-domain encoder in terms of having higher performance and efficiency, because it exploits the smoothness and sparsity of human organs. Experiments have been conducted on both human organ and phantom datasets. The proposed method achieves state-of-the-art accuracy and spends reasonably short computation time.

Hongguang Zhang, Limeng Zhang, Yuchao Dai et al.

Contemporary deep learning multi-scale deblurring models suffer from many issues: 1) They perform poorly on non-uniformly blurred images/videos; 2) Simply increasing the model depth with finer-scale levels cannot improve deblurring; 3) Individual RGB frames contain a limited motion information for deblurring; 4) Previous models have a limited robustness to spatial transformations and noise. Below, we extend the DMPHN model by several mechanisms to address the above issues: I) We present a novel self-supervised event-guided deep hierarchical Multi-patch Network (MPN) to deal with blurry images and videos via fine-to-coarse hierarchical localized representations; II) We propose a novel stacked pipeline, StackMPN, to improve the deblurring performance under the increased network depth; III) We propose an event-guided architecture to exploit motion cues contained in videos to tackle complex blur in videos; IV) We propose a novel self-supervised step to expose the model to random transformations (rotations, scale changes), and make it robust to Gaussian noises. Our MPN achieves the state of the art on the GoPro and VideoDeblur datasets with a 40x faster runtime compared to current multi-scale methods. With 30ms to process an image at 1280x720 resolution, it is the first real-time deep motion deblurring model for 720p images at 30fps. For StackMPN, we obtain significant improvements over 1.2dB on the GoPro dataset by increasing the network depth. Utilizing the event information and self-supervision further boost results to 33.83dB.

Shan Wang, Yanhao Zhang, Ankit Vora et al.

Existing spatial localization techniques for autonomous vehicles mostly use a pre-built 3D-HD map, often constructed using a survey-grade 3D mapping vehicle, which is not only expensive but also laborious. This paper shows that by using an off-the-shelf high-definition satellite image as a ready-to-use map, we are able to achieve cross-view vehicle localization up to a satisfactory accuracy, providing a cheaper and more practical way for localization. While the utilization of satellite imagery for cross-view localization is an established concept, the conventional methodology focuses primarily on image retrieval. This paper introduces a novel approach to cross-view localization that departs from the conventional image retrieval method. Specifically, our method develops (1) a Geometric-align Feature Extractor (GaFE) that leverages measured 3D points to bridge the geometric gap between ground and overhead views, (2) a Pose Aware Branch (PAB) adopting a triplet loss to encourage pose-aware feature extraction, and (3) a Recursive Pose Refine Branch (RPRB) using the Levenberg-Marquardt (LM) algorithm to align the initial pose towards the true vehicle pose iteratively. Our method is validated on KITTI and Ford Multi-AV Seasonal datasets as ground view and Google Maps as the satellite view. The results demonstrate the superiority of our method in cross-view localization with median spatial and angular errors within $1$ meter and $1^\circ$, respectively.

Yujiao Shi, Hongdong Li

This paper addresses the problem of vehicle-mounted camera localization by matching a ground-level image with an overhead-view satellite map. Existing methods often treat this problem as cross-view image retrieval, and use learned deep features to match the ground-level query image to a partition (eg, a small patch) of the satellite map. By these methods, the localization accuracy is limited by the partitioning density of the satellite map (often in the order of tens meters). Departing from the conventional wisdom of image retrieval, this paper presents a novel solution that can achieve highly-accurate localization. The key idea is to formulate the task as pose estimation and solve it by neural-net based optimization. Specifically, we design a two-branch {CNN} to extract robust features from the ground and satellite images, respectively. To bridge the vast cross-view domain gap, we resort to a Geometry Projection module that projects features from the satellite map to the ground-view, based on a relative camera pose. Aiming to minimize the differences between the projected features and the observed features, we employ a differentiable Levenberg-Marquardt ({LM}) module to search for the optimal camera pose iteratively. The entire pipeline is differentiable and runs end-to-end. Extensive experiments on standard autonomous vehicle localization datasets have confirmed the superiority of the proposed method. Notably, e.g., starting from a coarse estimate of camera location within a wide region of 40m x 40m, with an 80% likelihood our method quickly reduces the lateral location error to be within 5m on a new KITTI cross-view dataset.

Yujiao Shi, Xin Yu, Liu Liu et al.

We address the problem of ground-to-satellite image geo-localization, that is, estimating the camera latitude, longitude and orientation (azimuth angle) by matching a query image captured at the ground level against a large-scale database with geotagged satellite images. Our prior arts treat the above task as pure image retrieval by selecting the most similar satellite reference image matching the ground-level query image. However, such an approach often produces coarse location estimates because the geotag of the retrieved satellite image only corresponds to the image center while the ground camera can be located at any point within the image. To further consolidate our prior research findings, we present a novel geometry-aware geo-localization method. Our new method is able to achieve the fine-grained location of a query image, up to pixel size precision of the satellite image, once its coarse location and orientation have been determined. Moreover, we propose a new geometry-aware image retrieval pipeline to improve the coarse localization accuracy. Apart from a polar transform in our conference work, this new pipeline also maps satellite image pixels to the ground-level plane in the ground-view via a geometry-constrained projective transform to emphasize informative regions, such as road structures, for cross-view geo-localization. Extensive quantitative and qualitative experiments demonstrate the effectiveness of our newly proposed framework. We also significantly improve the performance of coarse localization results compared to the state-of-the-art in terms of location recalls.

Jinguang Tong, Sundaram Muthu, Fahira Afzal Maken et al.

In this paper, we define a new problem of recovering the 3D geometry of an object confined in a transparent enclosure. We also propose a novel method for solving this challenging problem. Transparent enclosures pose challenges of multiple light reflections and refractions at the interface between different propagation media e.g. air or glass. These multiple reflections and refractions cause serious image distortions which invalidate the single viewpoint assumption. Hence the 3D geometry of such objects cannot be reliably reconstructed using existing methods, such as traditional structure from motion or modern neural reconstruction methods. We solve this problem by explicitly modeling the scene as two distinct sub-spaces, inside and outside the transparent enclosure. We use an existing neural reconstruction method (NeuS) that implicitly represents the geometry and appearance of the inner subspace. In order to account for complex light interactions, we develop a hybrid rendering strategy that combines volume rendering with ray tracing. We then recover the underlying geometry and appearance of the model by minimizing the difference between the real and hybrid rendered images. We evaluate our method on both synthetic and real data. Experiment results show that our method outperforms the state-of-the-art (SOTA) methods. Codes and data will be available at https://github.com/hirotong/ReNeuS

Haorui Ji, Hui Deng, Yuchao Dai et al.

Most of the previous 3D human pose estimation work relied on the powerful memory capability of the network to obtain suitable 2D-3D mappings from the training data. Few works have studied the modeling of human posture deformation in motion. In this paper, we propose a new modeling method for human pose deformations and design an accompanying diffusion-based motion prior. Inspired by the field of non-rigid structure-from-motion, we divide the task of reconstructing 3D human skeletons in motion into the estimation of a 3D reference skeleton, and a frame-by-frame skeleton deformation. A mixed spatial-temporal NRSfMformer is used to simultaneously estimate the 3D reference skeleton and the skeleton deformation of each frame from 2D observations sequence, and then sum them to obtain the pose of each frame. Subsequently, a loss term based on the diffusion model is used to ensure that the pipeline learns the correct prior motion knowledge. Finally, we have evaluated our proposed method on mainstream datasets and obtained superior results outperforming the state-of-the-art.

Shan Wang, Yanhao Zhang, Akhil Perincherry et al.

This paper proposes a fine-grained self-localization method for outdoor robotics that utilizes a flexible number of onboard cameras and readily accessible satellite images. The proposed method addresses limitations in existing cross-view localization methods that struggle to handle noise sources such as moving objects and seasonal variations. It is the first sparse visual-only method that enhances perception in dynamic environments by detecting view-consistent key points and their corresponding deep features from ground and satellite views, while removing off-the-ground objects and establishing homography transformation between the two views. Moreover, the proposed method incorporates a spatial embedding approach that leverages camera intrinsic and extrinsic information to reduce the ambiguity of purely visual matching, leading to improved feature matching and overall pose estimation accuracy. The method exhibits strong generalization and is robust to environmental changes, requiring only geo-poses as ground truth. Extensive experiments on the KITTI and Ford Multi-AV Seasonal datasets demonstrate that our proposed method outperforms existing state-of-the-art methods, achieving median spatial accuracy errors below $0.5$ meters along the lateral and longitudinal directions, and a median orientation accuracy error below 2 degrees.

Bin Fan, Yuchao Dai, Hongdong Li

A single rolling-shutter (RS) image may be viewed as a row-wise combination of a sequence of global-shutter (GS) images captured by a (virtual) moving GS camera within the exposure duration. Although RS cameras are widely used, the RS effect causes obvious image distortion especially in the presence of fast camera motion, hindering downstream computer vision tasks. In this paper, we propose to invert the RS image capture mechanism, i.e., recovering a continuous high framerate GS video from two time-consecutive RS frames. We call this task the RS temporal super-resolution (RSSR) problem. The RSSR is a very challenging task, and to our knowledge, no practical solution exists to date. This paper presents a novel deep-learning based solution. By leveraging the multi-view geometry relationship of the RS imaging process, our learning-based framework successfully achieves high framerate GS generation. Specifically, three novel contributions can be identified: (i) novel formulations for bidirectional RS undistortion flows under constant velocity as well as constant acceleration motion model. (ii) a simple linear scaling operation, which bridges the RS undistortion flow and regular optical flow. (iii) a new mutual conversion scheme between varying RS undistortion flows that correspond to different scanlines. Our method also exploits the underlying spatial-temporal geometric relationships within a deep learning framework, where no additional supervision is required beyond the necessary middle-scanline GS image. Building upon these contributions, we represent the very first rolling-shutter temporal super-resolution deep-network that is able to recover high framerate GS videos from just two RS frames. Extensive experimental results on both synthetic and real data show that our proposed method can produce high-quality GS image sequences with rich details, outperforming the state-of-the-art methods.

Junxuan Li, Hongdong Li

This paper aims at recovering the shape of a scene with unknown, non-Lambertian, and possibly spatially-varying surface materials. When the shape of the object is highly complex and that shadows cast on the surface, the task becomes very challenging. To overcome these challenges, we propose a coordinate-based deep MLP (multilayer perceptron) to parameterize both the unknown 3D shape and the unknown reflectance at every surface point. This network is able to leverage the observed photometric variance and shadows on the surface, and recover both surface shape and general non-Lambertian reflectance. We explicitly predict cast shadows, mitigating possible artifacts on these shadowing regions, leading to higher estimation accuracy. Our framework is entirely self-supervised, in the sense that it requires neither ground truth shape nor BRDF. Tests on real-world images demonstrate that our method outperform existing methods by a significant margin. Thanks to the small size of the MLP-net, our method is an order of magnitude faster than previous CNN-based methods.

Hui Deng, Tong Zhang, Yuchao Dai et al.

Directly regressing the non-rigid shape and camera pose from the individual 2D frame is ill-suited to the Non-Rigid Structure-from-Motion (NRSfM) problem. This frame-by-frame 3D reconstruction pipeline overlooks the inherent spatial-temporal nature of NRSfM, i.e., reconstructing the whole 3D sequence from the input 2D sequence. In this paper, we propose to model deep NRSfM from a sequence-to-sequence translation perspective, where the input 2D frame sequence is taken as a whole to reconstruct the deforming 3D non-rigid shape sequence. First, we apply a shape-motion predictor to estimate the initial non-rigid shape and camera motion from a single frame. Then we propose a context modeling module to model camera motions and complex non-rigid shapes. To tackle the difficulty in enforcing the global structure constraint within the deep framework, we propose to impose the union-of-subspace structure by replacing the self-expressiveness layer with multi-head attention and delayed regularizers, which enables end-to-end batch-wise training. Experimental results across different datasets such as Human3.6M, CMU Mocap and InterHand prove the superiority of our framework.

Junxuan Li, Hongdong Li

This paper tackles the task of uncalibrated photometric stereo for 3D object reconstruction, where both the object shape, object reflectance, and lighting directions are unknown. This is an extremely difficult task, and the challenge is further compounded with the existence of the well-known generalized bas-relief (GBR) ambiguity in photometric stereo. Previous methods to resolve this ambiguity either rely on an overly simplified reflectance model, or assume special light distribution. We propose a new method that jointly optimizes object shape, light directions, and light intensities, all under general surfaces and lights assumptions. The specularities are used explicitly to solve uncalibrated photometric stereo via a neural inverse rendering process. We gradually fit specularities from shiny to rough using novel progressive specular bases. Our method leverages a physically based rendering equation by minimizing the reconstruction error on a per-object-basis. Our method demonstrates state-of-the-art accuracy in light estimation and shape recovery on real-world datasets.

Yujiao Shi, Xin Yu, Shan Wang et al.

This paper tackles the problem of Cross-view Video-based camera Localization (CVL). The task is to localize a query camera by leveraging information from its past observations, i.e., a continuous sequence of images observed at previous time stamps, and matching them to a large overhead-view satellite image. The critical challenge of this task is to learn a powerful global feature descriptor for the sequential ground-view images while considering its domain alignment with reference satellite images. For this purpose, we introduce CVLNet, which first projects the sequential ground-view images into an overhead view by exploring the ground-and-overhead geometric correspondences and then leverages the photo consistency among the projected images to form a global representation. In this way, the cross-view domain differences are bridged. Since the reference satellite images are usually pre-cropped and regularly sampled, there is always a misalignment between the query camera location and its matching satellite image center. Motivated by this, we propose estimating the query camera's relative displacement to a satellite image before similarity matching. In this displacement estimation process, we also consider the uncertainty of the camera location. For example, a camera is unlikely to be on top of trees. To evaluate the performance of the proposed method, we collect satellite images from Google Map for the KITTI dataset and construct a new cross-view video-based localization benchmark dataset, KITTI-CVL. Extensive experiments have demonstrated the effectiveness of video-based localization over single image-based localization and the superiority of each proposed module over other alternatives.

Ziang Cheng, Junxuan Li, Hongdong Li

This paper proposes a practical photometric solution for the challenging problem of in-the-wild inverse rendering under unknown ambient lighting. Our system recovers scene geometry and reflectance using only multi-view images captured by a smartphone. The key idea is to exploit smartphone's built-in flashlight as a minimally controlled light source, and decompose image intensities into two photometric components -- a static appearance corresponds to ambient flux, plus a dynamic reflection induced by the moving flashlight. Our method does not require flash/non-flash images to be captured in pairs. Building on the success of neural light fields, we use an off-the-shelf method to capture the ambient reflections, while the flashlight component enables physically accurate photometric constraints to decouple reflectance and illumination. Compared to existing inverse rendering methods, our setup is applicable to non-darkroom environments yet sidesteps the inherent difficulties of explicit solving ambient reflections. We demonstrate by extensive experiments that our method is easy to implement, casual to set up, and consistently outperforms existing in-the-wild inverse rendering techniques. Finally, our neural reconstruction can be easily exported to PBR textured triangle mesh ready for industrial renderers.

Junxuan Li, Shunsuke Saito, Tomas Simon et al.

Eyeglasses play an important role in the perception of identity. Authentic virtual representations of faces can benefit greatly from their inclusion. However, modeling the geometric and appearance interactions of glasses and the face of virtual representations of humans is challenging. Glasses and faces affect each other's geometry at their contact points, and also induce appearance changes due to light transport. Most existing approaches do not capture these physical interactions since they model eyeglasses and faces independently. Others attempt to resolve interactions as a 2D image synthesis problem and suffer from view and temporal inconsistencies. In this work, we propose a 3D compositional morphable model of eyeglasses that accurately incorporates high-fidelity geometric and photometric interaction effects. To support the large variation in eyeglass topology efficiently, we employ a hybrid representation that combines surface geometry and a volumetric representation. Unlike volumetric approaches, our model naturally retains correspondences across glasses, and hence explicit modification of geometry, such as lens insertion and frame deformation, is greatly simplified. In addition, our model is relightable under point lights and natural illumination, supporting high-fidelity rendering of various frame materials, including translucent plastic and metal within a single morphable model. Importantly, our approach models global light transport effects, such as casting shadows between faces and glasses. Our morphable model for eyeglasses can also be fit to novel glasses via inverse rendering. We compare our approach to state-of-the-art methods and demonstrate significant quality improvements.

Jianyuan Wang, Lalit Bhagat, Ceyuan Yang et al.

Generative models make huge progress to the photorealistic image synthesis in recent years. To enable human to steer the image generation process and customize the output, many works explore the interpretable dimensions of the latent space in GANs. Existing methods edit the attributes of the output image such as orientation or color scheme by varying the latent code along certain directions. However, these methods usually require additional human annotations for each pretrained model, and they mostly focus on editing global attributes. In this work, we propose a self-supervised approach to improve the spatial steerability of GANs without searching for steerable directions in the latent space or requiring extra annotations. Specifically, we design randomly sampled Gaussian heatmaps to be encoded into the intermediate layers of generative models as spatial inductive bias. Along with training the GAN model from scratch, these heatmaps are being aligned with the emerging attention of the GAN's discriminator in a self-supervised learning manner. During inference, users can interact with the spatial heatmaps in an intuitive manner, enabling them to edit the output image by adjusting the scene layout, moving, or removing objects. Moreover, we incorporate DragGAN into our framework, which facilitates fine-grained manipulation within a reasonable time and supports a coarse-to-fine editing process. Extensive experiments show that the proposed method not only enables spatial editing over human faces, animal faces, outdoor scenes, and complicated multi-object indoor scenes but also brings improvement in synthesis quality. Code, models, and demo video are available at https://genforce.github.io/SpatialGAN/.

Xuanxi Chen, Tao Wang, Ziqian Shao et al.

Images or videos captured by the Under-Display Camera (UDC) suffer from severe degradation, such as saturation degeneration and color shift. While restoration for UDC has been a critical task, existing works of UDC restoration focus only on images. UDC video restoration (UDC-VR) has not been explored in the community. In this work, we first propose a GAN-based generation pipeline to simulate the realistic UDC degradation process. With the pipeline, we build the first large-scale UDC video restoration dataset called PexelsUDC, which includes two subsets named PexelsUDC-T and PexelsUDC-P corresponding to different displays for UDC. Using the proposed dataset, we conduct extensive benchmark studies on existing video restoration methods and observe their limitations on the UDC-VR task. To this end, we propose a novel transformer-based baseline method that adaptively enhances degraded videos. The key components of the method are a spatial branch with local-aware transformers, a temporal branch embedded temporal transformers, and a spatial-temporal fusion module. These components drive the model to fully exploit spatial and temporal information for UDC-VR. Extensive experiments show that our method achieves state-of-the-art performance on PexelsUDC. The benchmark and the baseline method are expected to promote the progress of UDC-VR in the community, which will be made public.

Xin Huang, Qi Zhang, Ying Feng et al.

Representing visual signals with implicit coordinate-based neural networks, as an effective replacement of the traditional discrete signal representation, has gained considerable popularity in computer vision and graphics. In contrast to existing implicit neural representations which focus on modelling the scene only, this paper proposes a novel implicit camera model which represents the physical imaging process of a camera as a deep neural network. We demonstrate the power of this new implicit camera model on two inverse imaging tasks: i) generating all-in-focus photos, and ii) HDR imaging. Specifically, we devise an implicit blur generator and an implicit tone mapper to model the aperture and exposure of the camera's imaging process, respectively. Our implicit camera model is jointly learned together with implicit scene models under multi-focus stack and multi-exposure bracket supervision. We have demonstrated the effectiveness of our new model on a large number of test images and videos, producing accurate and visually appealing all-in-focus and high dynamic range images. In principle, our new implicit neural camera model has the potential to benefit a wide array of other inverse imaging tasks.

Xiaoxiao Sun, Nidham Gazagnadou, Vivek Sharma et al.

Hand-crafted image quality metrics, such as PSNR and SSIM, are commonly used to evaluate model privacy risk under reconstruction attacks. Under these metrics, reconstructed images that are determined to resemble the original one generally indicate more privacy leakage. Images determined as overall dissimilar, on the other hand, indicate higher robustness against attack. However, there is no guarantee that these metrics well reflect human opinions, which, as a judgement for model privacy leakage, are more trustworthy. In this paper, we comprehensively study the faithfulness of these hand-crafted metrics to human perception of privacy information from the reconstructed images. On 5 datasets ranging from natural images, faces, to fine-grained classes, we use 4 existing attack methods to reconstruct images from many different classification models and, for each reconstructed image, we ask multiple human annotators to assess whether this image is recognizable. Our studies reveal that the hand-crafted metrics only have a weak correlation with the human evaluation of privacy leakage and that even these metrics themselves often contradict each other. These observations suggest risks of current metrics in the community. To address this potential risk, we propose a learning-based measure called SemSim to evaluate the Semantic Similarity between the original and reconstructed images. SemSim is trained with a standard triplet loss, using an original image as an anchor, one of its recognizable reconstructed images as a positive sample, and an unrecognizable one as a negative. By training on human annotations, SemSim exhibits a greater reflection of privacy leakage on the semantic level. We show that SemSim has a significantly higher correlation with human judgment compared with existing metrics. Moreover, this strong correlation generalizes to unseen datasets, models and attack methods.

Rong Wang, Wei Mao, Hongdong Li

This paper addresses the task of 3D pose estimation for a hand interacting with an object from a single image observation. When modeling hand-object interaction, previous works mainly exploit proximity cues, while overlooking the dynamical nature that the hand must stably grasp the object to counteract gravity and thus preventing the object from slipping or falling. These works fail to leverage dynamical constraints in the estimation and consequently often produce unstable results. Meanwhile, refining unstable configurations with physics-based reasoning remains challenging, both by the complexity of contact dynamics and by the lack of effective and efficient physics inference in the data-driven learning framework. To address both issues, we present DeepSimHO: a novel deep-learning pipeline that combines forward physics simulation and backward gradient approximation with a neural network. Specifically, for an initial hand-object pose estimated by a base network, we forward it to a physics simulator to evaluate its stability. However, due to non-smooth contact geometry and penetration, existing differentiable simulators can not provide reliable state gradient. To remedy this, we further introduce a deep network to learn the stability evaluation process from the simulator, while smoothly approximating its gradient and thus enabling effective back-propagation. Extensive experiments show that our method noticeably improves the stability of the estimation and achieves superior efficiency over test-time optimization. The code is available at https://github.com/rongakowang/DeepSimHO.

Rong Wang, Wei Mao, Changsheng Lu et al.

Animating stylized characters to match a reference motion sequence is a highly demanded task in film and gaming industries. Existing methods mostly focus on rigid deformations of characters' body, neglecting local deformations on the apparel driven by physical dynamics. They deform apparel the same way as the body, leading to results with limited details and unrealistic artifacts, e.g. body-apparel penetration. In contrast, we present a novel method aiming for high-quality motion transfer with realistic apparel animation. As existing datasets lack annotations necessary for generating realistic apparel animations, we build a new dataset named MMDMC, which combines stylized characters from the MikuMikuDance community with real-world Motion Capture data. We then propose a data-driven pipeline that learns to disentangle body and apparel deformations via two neural deformation modules. For body parts, we propose a geodesic attention block to effectively incorporate semantic priors into skeletal body deformation to tackle complex body shapes for stylized characters. Since apparel motion can significantly deviate from respective body joints, we propose to model apparel deformation in a non-linear vertex displacement field conditioned on its historic states. Extensive experiments show that our method produces results with superior quality for various types of apparel. Our dataset is released in https://github.com/rongakowang/MMDMC.

Chushan Zhang, Ruihan Lu, Jinguang Tong et al.

Leveraging 3D information within Multimodal Large Language Models (MLLMs) has recently shown significant advantages for indoor scene understanding. However, existing methods, including those using explicit ground-truth 3D positional encoding and those grafting external 3D foundation models for implicit geometry, struggle with the trade-off in 2D-3D representation fusion, leading to suboptimal deployment. To this end, we propose 3D-Implicit Depth Emergence, a method that reframes 3D perception as an emergent property derived from geometric self-supervision rather than explicit encoding. Our core insight is the Implicit Geometric Emergence Principle: by strategically leveraging privileged geometric supervision through mechanisms like a fine-grained geometry validator and global representation constraints, we construct an information bottleneck. This bottleneck forces the model to maximize the mutual information between visual features and 3D structures, allowing 3D awareness to emerge naturally within a unified visual representation. Unlike existing approaches, our method enables 3D perception to emerge implicitly, disentangling features in dense regions and, crucially, eliminating depth and pose dependencies during inference with zero latency overhead. This paradigm shift from external grafting to implicit emergence represents a fundamental rethinking of 3D knowledge integration in visual-language models. Extensive experiments demonstrate that our method surpasses SOTA on multiple 3D scene understanding benchmarks. Our approach achieves a 55% reduction in inference latency while maintaining strong performance across diverse downstream tasks, underscoring the effectiveness of meticulously designed auxiliary objectives for dependency-free 3D understanding. Source code can be found at github.com/ChushanZhang/3D-IDE.

Ziang Cheng, Shihao Jiang, Hongdong Li

The problem of video frame interpolation is to increase the temporal resolution of a low frame-rate video, by interpolating novel frames between existing temporally sparse frames. This paper presents a self-supervised approach to video frame interpolation that requires only a single video. We pose the video as a set of layers. Each layer is parameterized by two implicit neural networks -- one for learning a static frame and the other for a time-varying motion field corresponding to video dynamics. Together they represent an occlusion-free subset of the scene with a pseudo-depth channel. To model inter-layer occlusions, all layers are lifted to the 2.5D space so that the frontal layer occludes distant layers. This is done by assigning each layer a depth channel, which we call `pseudo-depth', whose partial order defines the occlusion between layers. The pseudo-depths are converted to visibility values through a fully differentiable SoftMin function so that closer layers are more visible than layers in a distance. On the other hand, we parameterize the video motions by solving an ordinary differentiable equation (ODE) defined on a time-varying neural velocity field that guarantees valid motions. This implicit neural representation learns the video as a space-time continuum, allowing frame interpolation at any temporal resolution. We demonstrate the effectiveness of our method on real-world datasets, where our method achieves comparable performance to state-of-the-arts that require ground truth labels for training.

Yongzhi Xu, Yonhon Ng, Yifu Wang et al.

3D Content Generation is at the heart of many computer graphics applications, including video gaming, film-making, virtual and augmented reality, etc. This paper proposes a novel deep-learning based approach for automatically generating interactive and playable 3D game scenes, all from the user's casual prompts such as a hand-drawn sketch. Sketch-based input offers a natural, and convenient way to convey the user's design intention in the content creation process. To circumvent the data-deficient challenge in learning (i.e. the lack of large training data of 3D scenes), our method leverages a pre-trained 2D denoising diffusion model to generate a 2D image of the scene as the conceptual guidance. In this process, we adopt the isometric projection mode to factor out unknown camera poses while obtaining the scene layout. From the generated isometric image, we use a pre-trained image understanding method to segment the image into meaningful parts, such as off-ground objects, trees, and buildings, and extract the 2D scene layout. These segments and layouts are subsequently fed into a procedural content generation (PCG) engine, such as a 3D video game engine like Unity or Unreal, to create the 3D scene. The resulting 3D scene can be seamlessly integrated into a game development environment and is readily playable. Extensive tests demonstrate that our method can efficiently generate high-quality and interactive 3D game scenes with layouts that closely follow the user's intention.

Rong Wang, Ruyi Zha, Ziang Cheng et al.

We present a novel method for generating geometrically realistic and consistent orbital videos from a single image of an object. Existing video generation works mostly rely on pixel-wise attention to enforce view consistency across frames. However, such mechanism does not impose sufficient constraints for long-range extrapolation, e.g. rear-view synthesis, in which pixel correspondences to the input image are limited. Consequently, these works often fail to produce results with a plausible and coherent structure. To tackle this issue, we propose to leverage rich shape priors from a 3D foundational generative model as an auxiliary constraint, motivated by its capability of modeling realistic object shape distributions learned from large 3D asset corpora. Specifically, we prompt the video generation with two scales of latent features encoded by the 3D foundation model: (i) a denoised global latent vector as an overall structural guidance, and (ii) a set of latent images projected from volumetric features to provide view-dependent and fine-grained geometry details. In contrast to commonly used 2.5D representations such as depth or normal maps, these compact features can model complete object shapes, and help to improve inference efficiency by avoiding explicit mesh extraction. To achieve effective shape conditioning, we introduce a multi-scale 3D adapter to inject feature tokens to the base video model via cross-attention, which retains its capabilities from general video pretraining and enables a simple and model-agonistic fine-tuning process. Extensive experiments on multiple benchmarks show that our method achieves superior visual quality, shape realism and multi-view consistency compared to state-of-the-art methods, and robustly generalizes to complex camera trajectories and in-the-wild images.

Qi Zhang, Xin Huang, Ying Feng et al.

This paper studies the problem of view synthesis with certain amount of rotations from a pair of images, what we called stereo unstructured magnification. While the multi-plane image representation is well suited for view synthesis with depth invariant, how to generalize it to unstructured views remains a significant challenge. This is primarily due to the depth-dependency caused by camera frontal parallel representation. Here we propose a novel multiple homography image (MHI) representation, comprising of a set of scene planes with fixed normals and distances. A two-stage network is developed for novel view synthesis. Stage-1 is an MHI reconstruction module that predicts the MHIs and composites layered multi-normal images along the normal direction. Stage-2 is a normal-blending module to find blending weights. We also derive an angle-based cost to guide the blending of multi-normal images by exploiting per-normal geometry. Compared with the state-of-the-art methods, our method achieves superior performance for view synthesis qualitatively and quantitatively, especially for cases when the cameras undergo rotations.

Wen-Yan Lin, Siying Liu, Bing Tian Dai et al.

Most classifiers rely on discriminative boundaries that separate instances of each class from everything else. We argue that discriminative boundaries are counter-intuitive as they define semantics by what-they-are-not; and should be replaced by generative classifiers which define semantics by what-they-are. Unfortunately, generative classifiers are significantly less accurate. This may be caused by the tendency of generative models to focus on easy to model semantic generative factors and ignore non-semantic factors that are important but difficult to model. We propose a new generative model in which semantic factors are accommodated by shell theory's hierarchical generative process and non-semantic factors by an instance specific noise term. We use the model to develop a classification scheme which suppresses the impact of noise while preserving semantic cues. The result is a surprisingly accurate generative classifier, that takes the form of a modified nearest-neighbor algorithm; we term it distance classification. Unlike discriminative classifiers, a distance classifier: defines semantics by what-they-are; is amenable to incremental updates; and scales well with the number of classes.

Zhi Jin, Yuwei Qiu, Kaihao Zhang et al.

Recently, Transformer networks have demonstrated outstanding performance in the field of image restoration due to the global receptive field and adaptability to input. However, the quadratic computational complexity of Softmax-attention poses a significant limitation on its extensive application in image restoration tasks, particularly for high-resolution images. To tackle this challenge, we propose a novel variant of the Transformer. This variant leverages the Taylor expansion to approximate the Softmax-attention and utilizes the concept of norm-preserving mapping to approximate the remainder of the first-order Taylor expansion, resulting in a linear computational complexity. Moreover, we introduce a multi-branch architecture featuring multi-scale patch embedding into the proposed Transformer, which has four distinct advantages: 1) various sizes of the receptive field; 2) multi-level semantic information; 3) flexible shapes of the receptive field; 4) accelerated training and inference speed. Hence, the proposed model, named the second version of Taylor formula expansion-based Transformer (for short MB-TaylorFormer V2) has the capability to concurrently process coarse-to-fine features, capture long-distance pixel interactions with limited computational cost, and improve the approximation of the Taylor expansion remainder. Experimental results across diverse image restoration benchmarks demonstrate that MB-TaylorFormer V2 achieves state-of-the-art performance in multiple image restoration tasks, such as image dehazing, deraining, desnowing, motion deblurring, and denoising, with very little computational overhead. The source code is available at https://github.com/FVL2020/MB-TaylorFormerV2.

Shaoheng Xu, Chunyi Sun, Jihui Zhang et al.

Individualized head-related impulse responses (HRIRs) enable binaural rendering, but dense per-listener measurements are costly. We address HRIR spatial up-sampling from sparse per-listener measurements: given a few measured HRIRs for a listener, predict HRIRs at unmeasured target directions. Prior learning methods often work in the frequency domain, rely on minimum-phase assumptions or separate timing models, and use a fixed direction grid, which can degrade temporal fidelity and spatial continuity. We propose BiFormer3D, a time-domain, grid-free binaural Transformer for reconstructing HRIRs at arbitrary directions from sparse inputs. It uses sinusoidal spatial features, a Conv1D refinement module, and auxiliary interaural time difference (ITD) and interaural level difference (ILD) heads. On SONICOM, it improves normalized mean squared error (NMSE), cosine distance, and ITD/ILD errors over prior methods; ablations validate modules and show minimum-phase pre-processing is unnecessary.

Chengzeng Feng, Jiacheng Wei, Cheng Chen et al.

We propose Prim2Room, a novel framework for controllable room mesh generation leveraging 2D layout conditions and 3D primitive retrieval to facilitate precise 3D layout specification. Diverging from existing methods that lack control and precision, our approach allows for detailed customization of room-scale environments. To overcome the limitations of previous methods, we introduce an adaptive viewpoint selection algorithm that allows the system to generate the furniture texture and geometry from more favorable views than predefined camera trajectories. Additionally, we employ non-rigid depth registration to ensure alignment between generated objects and their corresponding primitive while allowing for shape variations to maintain diversity. Our method not only enhances the accuracy and aesthetic appeal of generated 3D scenes but also provides a user-friendly platform for detailed room design.

Yujiao Shi, Hongdong Li, Akhil Perincherry et al.

The ground-to-satellite image matching/retrieval was initially proposed for city-scale ground camera localization. This work addresses the problem of improving camera pose accuracy by ground-to-satellite image matching after a coarse location and orientation have been obtained, either from the city-scale retrieval or from consumer-level GPS and compass sensors. Existing learning-based methods for solving this task require accurate GPS labels of ground images for network training. However, obtaining such accurate GPS labels is difficult, often requiring an expensive {\color{black}Real Time Kinematics (RTK)} setup and suffering from signal occlusion, multi-path signal disruptions, \etc. To alleviate this issue, this paper proposes a weakly supervised learning strategy for ground-to-satellite image registration when only noisy pose labels for ground images are available for network training. It derives positive and negative satellite images for each ground image and leverages contrastive learning to learn feature representations for ground and satellite images useful for translation estimation. We also propose a self-supervision strategy for cross-view image relative rotation estimation, which trains the network by creating pseudo query and reference image pairs. Experimental results show that our weakly supervised learning strategy achieves the best performance on cross-area evaluation compared to recent state-of-the-art methods that are reliant on accurate pose labels for supervision.

Chushan Zhang, Ruihan Lu, Jinguang Tong et al.

Chunked vision-language-action (VLA) policies predict multi-step robot controls, conditioning each update on the current visual observation alone. Yet robot actions cause contact, occlusion, and object motion, and the geometry that later decisions depend on can change before the next visual update arrives. Spatial VLAs improve current-frame geometry. Temporal VLAs aggregate past frames. Neither maintains an action-updated scene prior across chunks. We argue for a persistent action-updated scene state across control calls, and introduce EvoScene-VLA. Its recurrent scene prefix carries a geometry-aware scene state across chunks. At each vision-language model (VLM) call, the VLM combines scene information from the current observation with the action-updated prior from the previous chunk; the action decoder outputs both the next action chunk and a compact scene update. This update becomes the next prior, which the VLM corrects against the new observation when the next call arrives. Each control call therefore starts from a scene prior that reflects both recent actions and fresh visual evidence. During training, \textbf{Scene Predictor} supplies future scene-token targets, and Geometric Anchor aligns scene slots with frozen depth and 3D teachers. We discard both modules at deployment. On 31 RoboTwin tasks, EvoScene-VLA raises average success from 87.2% to 89.1% in fixed evaluation and from 86.1% to 88.5% in randomized evaluation. On the Galaxea R1-Lite real robot, EvoScene-VLA outperforms all baselines.

Ziang Cheng, Jiayu Yang, Hongdong Li

Real-time Stereo Matching is a cornerstone algorithm for many Extended Reality (XR) applications, such as indoor 3D understanding, video pass-through, and mixed-reality games. Despite significant advancements in deep stereo methods, achieving real-time depth inference with high accuracy on a low-power device remains a major challenge. One of the major difficulties is the lack of high-quality indoor video stereo training datasets captured by head-mounted VR/AR glasses. To address this issue, we introduce a novel video stereo synthetic dataset that comprises photorealistic renderings of various indoor scenes and realistic camera motion captured by a 6-DoF moving VR/AR head-mounted display (HMD). This facilitates the evaluation of existing approaches and promotes further research on indoor augmented reality scenarios. Our newly proposed dataset enables us to develop a novel framework for continuous video-rate stereo matching. As another contribution, our dataset enables us to proposed a new video-based stereo matching approach tailored for XR applications, which achieves real-time inference at an impressive 134fps on a standard desktop computer, or 30fps on a battery-powered HMD. Our key insight is that disparity and contextual information are highly correlated and redundant between consecutive stereo frames. By unrolling an iterative cost aggregation in time (i.e. in the temporal dimension), we are able to distribute and reuse the aggregated features over time. This approach leads to a substantial reduction in computation without sacrificing accuracy. We conducted extensive evaluations and comparisons and demonstrated that our method achieves superior performance compared to the current state-of-the-art, making it a strong contender for real-time stereo matching in VR/AR applications.

Shan Wang, Peixia Li, Chenchen Xu et al.

We propose Light-Geometry Interaction (LGI) maps, a novel representation that encodes light-aware occlusion from monocular depth. Unlike ray tracing, which requires full 3D reconstruction, LGI captures essential light-shadow interactions reliably and accurately, computed from off-the-shelf 2.5D depth map predictions. LGI explicitly ties illumination direction to geometry, providing a physics-inspired prior that constrains generative models. Without such prior, these models often produce floating shadows, inconsistent illumination, and implausible shadow geometry. Building on this representation, we propose a unified pipeline for joint shadow generation and relighting - unlike prior methods that treat them as disjoint tasks - capturing the intrinsic coupling of illumination and shadowing essential for modeling indirect effects. By embedding LGI into a bridge-matching generative backbone, we reduce ambiguity and enforce physically consistent light-shadow reasoning. To enable effective training, we curated the first large-scale benchmark dataset for joint shadow and relighting, covering reflections, transparency, and complex interreflections. Experiments show significant gains in realism and consistency across synthetic and real images. LGI thus bridges geometry-inspired rendering with generative modeling, enabling efficient, physically consistent shadow generation and relighting.

Shiyan Su, Ruyi Zha, Danli Shi et al.

Neural representations (NRs), such as neural fields and 3D Gaussians, effectively model volumetric data in computed tomography (CT) but suffer from severe artifacts under sparse-view settings. To address this, we propose DiffNR, a novel framework that enhances NR optimization with diffusion priors. At its core is SliceFixer, a single-step diffusion model designed to correct artifacts in degraded slices. We integrate specialized conditioning layers into the network and develop tailored data curation strategies to support model finetuning. During reconstruction, SliceFixer periodically generates pseudo-reference volumes, providing auxiliary 3D perceptual supervision to fix underconstrained regions. Compared to prior methods that embed CT solvers into time-consuming iterative denoising, our repair-and-augment strategy avoids frequent diffusion model queries, leading to better runtime performance. Extensive experiments show that DiffNR improves PSNR by 3.99 dB on average, generalizes well across domains, and maintains efficient optimization.

Jinguang Tong, Jinbo Wu, Kaisiyuan Wang et al.

Human-Object Interaction (HOI) video reenactment with realistic motion remains a frontier in expressive digital human creation. Existing approaches primarily handle simple image-plane motion (e.g., in-plane translations), struggling with complex non-planar manipulations like out-of-plane reorientation. In this paper, we propose MVHOI, a two-stage HOI video reenactment framework that bridges multi-view reference conditions and video foundation models via a 3D Foundation Model (3DFM). The 3DFM first produces view-consistent object priors conditioned on implicit motion dynamics across novel viewpoints. A controllable video generation model then synthesizes high-fidelity object texture by incorporating multi-view reference images, ensuring appearance consistency via a reasonable retrieval mechanism. By enabling these two stages to mutually reinforce one another during the inference phase, our framework shows superior performance in generating long-duration HOI videos with intricate object manipulations. Extensive experiments show substantial improvements over prior approaches, especially for HOI with complex 3D object manipulations.

Yasong Dai, Zeeshan Hayder, David Ahmedt-Aristizabal et al.

Recent diffusion and flow matching models have demonstrated strong capabilities in image generation and editing by progressively removing noise through iterative sampling. While this enables flexible inversion for semantic-preserving edits, few-step sampling regimes suffer from poor forward process approximation, leading to degraded editing quality. Existing few-step inversion methods often rely on pretrained generators and auxiliary modules, limiting scalability and generalization across different architectures. To address these limitations, we propose BiFM (Bidirectional Flow Matching), a unified framework that jointly learns generation and inversion within a single model. BiFM directly estimates average velocity fields in both ``image $\to$ noise" and ``noise $\to$ image" directions, constrained by a shared instantaneous velocity field derived from either predefined schedules or pretrained multi-step diffusion models. Additionally, BiFM introduces a novel training strategy using continuous time-interval supervision, stabilized by a bidirectional consistency objective and a lightweight time-interval embedding. This bidirectional formulation also enables one-step inversion and can integrate seamlessly into popular diffusion and flow matching backbones. Across diverse image editing and generation tasks, BiFM consistently outperforms existing few-step approaches, achieving superior performance and editability.

Jin Liu, Xiaokang Pan, Junwen Duan et al.

This paper delves into the realm of stochastic optimization for compositional minimax optimization - a pivotal challenge across various machine learning domains, including deep AUC and reinforcement learning policy evaluation. Despite its significance, the problem of compositional minimax optimization is still under-explored. Adding to the complexity, current methods of compositional minimax optimization are plagued by sub-optimal complexities or heavy reliance on sizable batch sizes. To respond to these constraints, this paper introduces a novel method, called Nested STOchastic Recursive Momentum (NSTORM), which can achieve the optimal sample complexity of $O(κ^3 /ε^3 )$ to obtain the $ε$-accuracy solution. We also demonstrate that NSTORM can achieve the same sample complexity under the Polyak-Łojasiewicz (PL)-condition - an insightful extension of its capabilities. Yet, NSTORM encounters an issue with its requirement for low learning rates, potentially constraining its real-world applicability in machine learning. To overcome this hurdle, we present ADAptive NSTORM (ADA-NSTORM) with adaptive learning rates. We demonstrate that ADA-NSTORM can achieve the same sample complexity but the experimental results show its more effectiveness. All the proposed complexities indicate that our proposed methods can match lower bounds to existing minimax optimizations, without requiring a large batch size in each iteration. Extensive experiments support the efficiency of our proposed methods.