Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Chao XuSuggested from co-authorship
- Katsushi IkeuchiSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works316 from public data
- DAFL: Data-Free Learning of Student Networks453
A novel framework for training efficient deep neural networks by exploiting generative adversarial networks (GANs) is proposed, where the pre-trained teacher networks are regarded as a fixed discriminator and the generator is utilized for derivating training samples which can obtain the maximum response on the discriminator.
- Data-Free Learning of Student Networks453
A novel framework for training efficient deep neural networks by exploiting generative adversarial networks (GANs) is proposed, where the pre-trained teacher networks are regarded as a fixed discriminator and the generator is utilized for derivating training samples which can obtain the maximum response on the discriminator.
- Real-Time Intermediate Flow Estimation for Video Frame Interpolation392
RIFE uses a neural network named IFNet that can estimate the intermediate flows end-to-end with much faster speed and can support arbitrary-timestep frame interpolation with the temporal encoding input.
- Robust Photometric Stereo via Low-Rank Matrix Completion and Recovery344
This work presents a new approach to robustly solve photometric stereo problems by using advanced convex optimization techniques that are guaranteed to find the correct low-rank matrix by simultaneously fixing its missing and erroneous entries.
- DIST: Rendering Deep Implicit Signed Distance Function With Differentiable Sphere Tracing315
This work proposes a differentiable sphere tracing algorithm that can effectively reconstruct accurate 3D shapes from various inputs, such as sparse depth and multi-view images, through inverse optimization and shows excellent generalization capability and robustness against various noises.
- 284
- CARS: Continuous Evolution for Efficient Neural Architecture Search277
This work develops an efficient continuous evolutionary approach for searching neural networks that provides a series of networks with the number of parameters ranging from 3.7M to 5.1M under mobile settings and surpasses those produced by the state-of-the-art methods on the benchmark ImageNet dataset.
- AdderNet: Do We Really Need Multiplications in Deep Learning?240
This paper develops a special back-propagation approach for AdderNets by investigating the full-precision gradient, and proposes an adaptive learning rate strategy to enhance the training procedure of Ad DerNets according to the magnitude of each neuron's gradient.
- Polarized 3D: High-Quality Depth Sensing with Polarization Cues220
A framework to combine surface normals from polarization (hereafter polarization normals) with an aligned depth map to overcome physics-based artifacts, allowing the benefits of polarization to be used to enhance depth maps.
- 209
- Deep Shape from Polarization157
This paper makes a first attempt to bring the Shape from Polarization (SfP) problem to the realm of deep learning, and achieves results that exceed the previous state-of-the-art on a challenging dataset.
- 155
- Self-Calibrating Deep Photometric Stereo Networks149
An uncalibrated photometric stereo method for non-Lambertian scenes based on deep learning that can effectively take advantage of intermediate supervision, resulting in reduced learning difficulty compared to a single-stage model is proposed.
- 142
- RIFE: Real-Time Intermediate Flow Estimation for Video Frame Interpolation.140
A real-time intermediate flow estimation algorithm (RIFE) for video frame interpolation (VFI) that can be trained end-to-end and achieve excellent performance and achieves state-of-the-art index on several benchmarks is proposed.
- CRRN: Multi-scale Guided Concurrent Reflection Removal Network137
This paper proposes the Concurrent Reflection Removal Network (CRRN), a network that integrates image appearance information and multi-scale gradient information with human perception inspired loss function, and is trained on a new dataset with 3250 reflection images taken under diverse real-world scenes.
- 135
- 134
- 130
- Occluded Imaging with Time-of-Flight Sensors126
It is concluded that off-the-shelf TOF cameras can look around corners and that performance depends on two primary factors: camera modulation frequency and the width of the specular lobe (“shininess”) of the wall.
- SSF-CNN: Spatial and Spectral Fusion with CNN for Hyperspectral Image Super-Resolution121
Experimental results validate that the proposed spatial and spectral fusion CNNs outperforms the state-of-the-art methods and baseline CNN architectures in both quantitative values and visual qualities.
- 1000× Faster Camera and Machine Vision with Ordinary Devices116
A spiking neural network-based machine vision system that combines the speed of the machine and the mechanism of biological vision, achieving high-speed object detection and tracking 1,000x faster than human vision is developed.
- 115
- 114
- 109
- 106
- Joint Filtering of Intensity Images and Neuromorphic Events for High-Resolution Noise-Robust Imaging105
- Multi-View Photometric Stereo: A Robust Solution and Benchmark Dataset for Spatially Varying Isotropic Materials103
A method to capture both 3D shape and spatially varying reflectance with a multi-view photometric stereo technique that works for general isotropic materials and quantitatively evaluate state-of-the-art MVPS on a newly collected benchmark dataset is presented.
- 100
- ChipGAN97
The advantages in visual quality compared with state-of-the-art networks and high stylization perceptual study scores show the effectiveness of the proposed ChipGAN, an end-to-end Generative Adversarial Network based architecture for photo to Chinese ink wash painting style transfer.
- Deep Photometric Stereo for Non-Lambertian Surfaces94
This paper addresses the problem of photometric stereo, in both calibrated and uncalibrated scenarios, for non-Lambertian surfaces based on deep learning with a fully convolutional deep network, which is able to handle surfaces with general and unknown isotropic reflectance.
- Distilling Portable Generative Adversarial Networks for Image Translation89
Qualitative and quantitative analysis by conducting experiments on benchmark datasets demonstrate that the proposed method can learn portable generative models with strong performance.
- Unbounded High Dynamic Range Photography Using a Modulo Camera88
It is shown that with limited bit depth, very high radiance levels can be recovered from a single modulus image with the newly proposed unwrapping algorithm for natural images.
- SPLINE-Net: Sparse Photometric Stereo Through Lighting Interpolation and Normal Estimation Networks84
SPLINE-Net is verified to outperform existing methods for photometric stereo of general BRDFs by using only ten images of different lights instead of using nearly one hundred images.
- 82
- Depth Sensing Using Geometrically Constrained Polarization Normals82
This work uses additional geometric information, from coarse depth maps, to constrain the shape information from polarization cues, and proposes a framework that combines surface normals from polarization (hereafter polarization normals) with an aligned depth map.
- NormAttention-PSN: A High-frequency Region Enhanced Photometric Stereo Network with Normalized Attention79
An attention-weighted loss is presented to produce better surface reconstructions, which applies a higher weight to the detail-preserving gradient loss in high-frequency areas, and a double-gate normalization method for non-Lambertian surfaces is adopted.
- 79
- 79
- $$\mathrm {CT^2}$$: Colorization Transformer via Color Tokens78
This paper proposes an end-to-end transformer-based model to overcome the challenges of automatic image colorization and introduces color tokens into the approach and treats the colorization task as a classification problem, which increases the saturation of results.
- Reflection Separation using a Pair of Unpolarized and Polarized Images78
This paper proposes to solve semireflector orientation estimation first to make the physical image formation well-posed and then learn to reliably separate two layers using a refinement network with gradient loss.
- FHDe2Net: Full High Definition Demoireing Network76
The proposed Full High Definition Demoir´eing Network (FHDe 2 Net) consists of a global to local cascaded removal branch to eradicate multi-scale moir´e patterns and a frequency based high-resolution content separation branch to retain details.
- A Semi-Supervised Assessor of Neural Architectures75
This paper employs an auto-encoder to discover meaningful representations of neural architectures and introduces a graph convolutional neural network to predict the performance of architectures based on the learned representations and their relation modeled by the graph.
- Deep Learning Methods for Calibrated Photometric Stereo and Beyond72
This paper provides a comprehensive review of existing deep learning-based calibrated photometric stereo methods utilizing orthographic cameras and directional light sources and summarizes the performance of deep learning photometric stereo models on the most widely-used benchmark data set.
- 68
- 68
- Optical Flow Estimation for Spiking Camera64
SCFlow, a tailored deep learning pipeline to estimate optical flow in high-speed scenes from spike streams which can adaptively remove the motion blur in spike streams according to the prior motion is presented.
- L-CoDe:Language-Based Colorization Using Color-Object Decoupled Conditions64
A predictor for object-color corresponding matrix (OCCM) and a novel attention transfer module (ATM) are introduced to solve the color-object coupling problem and Experimental results show that the approach outperforms state-of-the-art methods conditioned on captions.
- 61
- 61
- Elevation Angle from Reflectance Monotonicity: Photometric Stereo for General Isotropic Reflectances60
The monotonicity of general isotropic reflectances is exploited for estimating elevation angles of surface normal given the azimuth angles to prove that elevation angles can be uniquely determined when the surface is observed under varying directional lights densely and uniformly distributed over the hemisphere.
- GPS-Net: Graph-based Photometric Stereo Network59
A Graph-based Photometric Stereo Network, which proposes the Unstructured Feature Extraction Layer to connect an arbitrary number of input image-light pairs into graph structures, and introduces Structure-aware Graph Convolution to balance the input data by appropriately weighting shadows and specular highlights.
- 58
- Bilateral Normal Integration55
This paper proposes a bilaterally weighted functional for discontinuity preserving normal integration and effectively preserves discontinuities and alleviates the under-or over-segmentation artifacts in the recovered surfaces compared to existing methods.
- 54
- 54
- On Positive-Unlabeled Classification in GAN54
A new Positive-Unlabeled GAN is derived that can achieve comparable or even better performance than those sophisticated discriminator stabilization methods and theoretically discusses the global optimality the proposed model will achieve and the equivalent optimization goal.
- 53
- 52
- Deblurring Low-Light Images with Events49
A unified two-stage framework along with a motion-aware neural network tailored to it is proposed, reconstructing the sharp image under the guidance of high-fidelity motion clues extracted from events robustly, and experimental results show the method achieves state-of-the-art performance on both synthetic and real-world images.
- 49
- 49
- Benchmarking Single-Image Reflection Removal Algorithms46
This paper presents a SIngle-image Reflection Removal Plus dataset, an up-to-date taxonomy for existing methods, a benchmark dataset, and the unified benchmarking evaluations for state-of-the-art (especially learning-based) methods.
- 46
- 43
- 42
- Polarization-Aware Low-Light Image Enhancement42
This paper proposes a Stokes-domain enhancement pipeline along with a dual-branch neural network to handle the problem in a polarization-aware manner and two application scenarios are presented to show how the enhancement can improve their results.
- Beyond Dropout: Feature Map Distortion to Regularize Deep Neural Networks42
This work investigates the empirical Rademacher complexity related to intermediate layers of deep neural networks and proposes a feature distortion method for addressing the problem of over-fitting.
- 41
- 41
- Summary study of data-driven photometric stereo methods41
Data-driven photometric stereo methods have shown that they possess a superior performance advantage over traditional methods, however, these methods suffer from various limitations, such as limited generalization capability.
- L-CoDer: Language-Based Colorization with Color-Object Decoupling Transformer40
This work introduces transformer into language-based colorization to tackle the aforementioned issues while keeping the language decoupling property, and unifies the modalities of image and language, and further performs color conditions evolving with image features in a coarse-to-fine manner.
- 40
- 39
- Context-Aware Cross-Attention for Skeleton-Based Human Action Recognition39
A cross-attention module that consists of a self-att attention branch and a cross-Attention branch for skeleton-based action recognition to extract joints that are not only more informative but also highly correlated to the corresponding scenario context information.
- Event-Driven Video Frame Synthesis39
A differentiable fusion model to approximate the dual-modal physical sensing process, unifying a variety of TVFS scenarios, e.g., interpolation, prediction and motion deblur is introduced and a deep learning strategy is developed to enhance the results from the first step, which is referred as a residual "denoising" process.
- LegoNet: Efficient Convolutional Neural Networks with Lego Filters39
A split-transform-merge strategy for an efficient convolution by exploiting intermediate Lego feature maps is developed and Inspired by network engineering, it is suggested that an ordinary filter in the neural network can be upgraded to a sophisticated module as well.
- Robust Student Network Learning39
This paper makes the student network produce more confident predictions with the help of the teacher network, and analyzes the lower bound of the perturbation that will destroy the confidence of theStudent network.
- GR-PSN: Learning to Estimate Surface Normal and Reconstruct Photometric Stereo Images38
A novel method, namely GR-PSN, which learns surface normals from photometric stereo images and generates the photometric images under distant illumination from different lighting directions and surface materials and shows realistic rendering results on 100 different materials is proposed.
- Face Image Reflection Removal38
The face image reflection removal problem is solved by incorporating inpainting ideas into a guided reflection removal framework, which takes two images as the input and considers various face-specific priors.
- 38
- 37
- 36
- 35
- 34
- SpecTrans: Versatile Material Classification for Interaction with Textureless, Specular and Transparent Surfaces34
This work presents SpecTrans, a new sensing technology for surface classification of exotic materials, such as glass, transparent plastic, and metal that requires significantly lower computational cost than conventional image-based methods, thereby providing real-time performance for ubiquitous computing.
- 33
- Reborn Filters: Pruning Convolutional Neural Networks with Limited Data33
This paper proposes to use all original filters to directly develop new compact filters, named reborn filters, so that all useful structure priors in the original filters can be well preserved into the pruned networks, alleviating the performance drop accordingly.
- 33
- 33
- 32
- 32
- Zero-Shot Low-Light Image Enhancement via Latent Diffusion Models31
A novel framework is presented that reformulates low-light image enhancement as a zero-shot inference problem using pre-trained latent diffusion models (LDMs), eliminating the need for task-specific training data and outperforms existing zero-shot methods across diverse real-world scenarios.
- Language-guided Image Reflection Separation31
A unified framework is proposed to solve the problem of language-guided re-flection separation, which leverages the cross-attention mechanism with contrastive learning strategies to construct the correspondence between language descriptions and image layers.
- Real-Time 3D-Aware Portrait Video Relighting31
This paper presents the first real-time 3D-aware method for relighting in-the-wild videos of talking faces based on Neural Radiance Fields (NeRF), and achieves state-of-the-art results in terms of reconstruction quality, lighting error, lighting instability, temporal consistency and inference speed.
- 30
- Deep Photometric Stereo Networks for Determining Surface Normal and Reflectances30
A deep photometric stereo network (DPSN) is proposed that takes reflectance observations under varying light directions and infers the surface normal and reflectance in a per-pixel manner and is applicable to real-world scenes.
- 29
- 29
- 29
- NTIRE 2025 Challenge on Event-Based Image Deblurring: Methods and Results28
This paper presents an overview of NTIRE 2025, the First Challenge on Event-Based Image Deblurring, detailing the proposed methodologies and corresponding results, and anticipate that this challenge will drive further advancements in event-based vision research.
- NeRSP: Neural 3D Reconstruction for Reflective Objects with Sparse Polarized Images28
This paper jointly han-dles the challenges from sparse inputs and reflective surfaces by leveraging polarized images and derives photomet-ric and geometric cues from the polarimetric image formation model and multiview azimuth consistency, which jointly optimize the surface geometry modeled via implicit neural representation.
- STAGE: Storyboard-Anchored Generation for Cinematic Multi-shot Narrative27
STAGE, a SToryboard-Anchored GEneration workflow to reformulate the keyframe-based multi-shot video generation task, and is introduced, a SToryboard-Anchored GEneration workflow to reformulate the keyframe-based multi-shot video generation task.
- Revisiting One-Stage Deep Uncalibrated Photometric Stereo via Fourier Embedding27
A one-stage deep uncalibrated photometric stereo (UPS) network, namely Fourier Uncalibrated Photometric Stereo Network (FUPS-Net), is introduced, offering better training stability, a concise end-to-end structure, and avoiding accumulated errors in disjointed networks.
- NEST: Neural Event Stack for Event-Based Image Enhancement27
A novel event representation named Neural Event STack is presented, which satisfies physical constraints and encodes comprehensive motion and temporal information sufficient for image enhancement and achieves superior performance on image deblurring and image super-resolution than state-of-the-art methods on both synthetic and real datasets.
- 27
- UnModNet: Learning to Unwrap a Modulo Image for High Dynamic Range Imaging27
This paper reformulates the modulo image unwrapping problem into a series of binary labeling problems and proposes a modulo edge-aware model, named as UnModNet, to iteratively estimate the binary rollover masks of the modulo image for unwrapping.
- EvHandPose: Event-Based 3D Hand Pose Estimation With Sparse Supervision26
The proposed EvHandPose with novel hand flow representations in Event-to-Pose module achieves accurate and stable hand pose estimation with high temporal resolution in fast motion and strong light scenes compared with RGB-based methods, generalizes well to outdoor scenes and another type of event camera, and shows the potential for the hand gesture recognition task.
- 26
- 26
- 26
- 26
- 25
- 25
- Spatially-Varying Outdoor Lighting Estimation from Intrinsics25
This work trains a deep neural network to regress intrinsic cues with physically-based constraints and use them to conduct global and local lightings estimation, and introduces the SOLID-Img dataset, a neural network for spatially- varying outdoor lighting estimation from a single outdoor image for any 2D pixel location.
- 25
- 25
- PanoWan: Lifting Diffusion Video Generation Models to 360° with Latitude/Longitude-aware Mechanisms24
PanoWan is introduced to effectively lift pre-trained text-to-video models to the panoramic domain, equipped with minimal modules, and achieves state-of-the-art performance in panoramic video generation and demonstrates robustness for zero-shot downstream tasks.
- Complementing Event Streams and RGB Frames for Hand Mesh Reconstruction24
EvRGBHand is proposed - the first approach for 3D hand mesh reconstruction with an event camera and an RGB camera compensating for each other, which can tackle overexposure and motion blur issues in RGB-based HMR and foreground scarcity as well as background overflow issues in event-based HMR.
- 24
- Neural Transmitted Radiance Fields24
This paper aims at addressing the problem of rendering novel transmitted views given a set of reflection-corrupted images by introducing the transmission encoder and recurring edge constraints as guidance, and can resist such reflection interference during rendering and reconstruct high-fidelity results even under sparse views.
- 23
- 23
- 23
- 22
- 22
- 21
- The NeRF Signature: Codebook-Aided Watermarking for Neural Radiance Fields20
This paper proposes NeRF Signature, a novel watermarking method for NeRF that employs a Codebook-aided Signature Embedding that does not alter the model structure, thereby maintaining imperceptibility and enhancing robustness at the model level.
- 20
- 20
- 20
- 20
- Video2Act: A Dual-System Video Diffusion Policy with Robotic Spatio-Motional Modeling19
Video2Act is a framework that efficiently guides robotic action learning by explicitly integrating spatial and motion-aware representations and surpasses previous state-of-the-art VLA methods by 7.7% in simulation and 21.7% in real-world tasks in terms of average success rate, further exhibiting strong generalization capabilities.
- DANI-Net: Uncalibrated Photometric Stereo by Differentiable Shadow Handling, Anisotropic Reflectance Modeling, and Neural Inverse Rendering19
DANI-Net is proposed, an inverse rendering framework with differentiable shadow handling and anisotropic reflectance modeling, which benefits from cues of shadow and reflectance through two differentiable paths.
- 19
- 19
- 19
- 19
- 18
- L-DiffER: Single Image Reflection Removal with Language-Based Diffusion Model17
This paper introduces L-DiffER, a language-based diffusion model designed for the ill-posed single image reflection removal task, and proposes an iterative condition refinement strategy to resolve the problem of inaccurate control conditions.
- 17
- Saliency detection by forward and backward cues in deep-CNN17
This paper proposes a top-down saliency model using CNN, a weakly supervised CNN model trained for 1000 object labelling task from RGB images that gives promising results compared to state-of-the-art models in two different datasets.
- 17
- 16
- 16
- 16
- Physics-based Neural Networks for Shape from Polarization.16
This work study the blending of physics and deep learning in the context of Shape from Polarization (SfP) in the framework of a two-stage encoder, finding that there is a subtlety to combining physics and neural networks.
- DeepShoe: An improved Multi-Task View-invariant CNN for street-to-shop shoe retrieval15
An improved Multi-Task View-invariant Convolutional Neural Network (MTV-CNN+) is proposed to handle the large visual discrepancy for the same shoe in different scenarios and an attribute-based weighting scheme on the conventional triplet loss function to put more emphasis on the hard triplets is developed.
- Sub-pixel Layout for Super-Resolution with Images in the Octic Group15
It is shown that recording multiple images, transformed in the octic group, with a sensor of asymmetric sub-pixel layout increases the spatial sampling compared to a conventional sensor with a rectilinear grid of pixels and hence increases the image resolution.
- 14
- 14
- SweepEvGS: Event-Based 3D Gaussian Splatting for Macro and Micro Radiance Field Rendering From a Single Sweep13
The results demonstrate that SweepEvGS surpasses existing methods in visual rendering quality, rendering speed, and computational efficiency, highlighting its potential for dynamic practical applications.
- Zero-Shot Event-Intensity Asymmetric Stereo via Visual Prompting from Image Domain13
This work proposes a zero-shot framework that utilizes monocular depth estimation and stereo matching models pretrained on diverse image datasets, and introduces a monocular cue-guided disparity refinement module to improve robustness across static and dynamic regions by incorporating monocular depth information from foundation models.
- 13
- DeRenderNet: Intrinsic Image Decomposition of Urban Scenes with Shape-(In)dependent Shading Rendering13
Compared with state-of-the-art intrinsic image decomposition methods, DeRenderNet produces shadow-free albedo maps with clean details and an accurate prediction of shadows in the shape-independent shading, which is shown to be effective in re- rendering and improving the accuracy of high-level vision tasks for urban scenes.
- 13
- 13
- 13
- 12
- Learning to Deblur Polarized Images12
A polarized image deblurring pipeline to solve the problem in a polarization-aware manner by adopting a divide-and-conquer strategy to explicitly decompose the problem into two less ill-posed sub-problems, and design a two-stage neural network to handle the two sub-problems respectively.
- Affective Image Editing: Shaping Emotional Factors via Text Descriptions12
This paper introduces AIEdiT for Affective Image Editing using Text descriptions, which evokes specific emotions by adaptively shaping multiple emotional factors across the entire image, and introduces an MLLM to supervise the model training.
- Optical design and characterization of an advanced computational imaging system12
An advanced computational imaging system with an optical architecture that enables simultaneous and dynamic pupil-plane and image-plane coding accommodating several task-specific applications is described.
- Reconstructing Satellites in 3D From Amateur Telescope Images11
This paper proposes a novel computational imaging framework that overcomes obstacles by integrating a hybrid image pre-processing pipeline with a joint pose estimation and 3D reconstruction module based on controlled Gaussian Splatting (GS) and Branch-and-Bound (BnB) search.
- Light Flickering Guided Reflection Removal11
This paper proposes a learning-based method that utilizes short-term and long-term observations of mixture videos to exploit one-side contextual clues in fluctuant components and brightness-consistent clues in consistent components for achieving layer separation and flickering removal, respectively.
- Learning from Bad Data via Generation11
Experimental results on different kinds of bad training data demonstrate the necessity and effectiveness of the proposed method and the influence of this generation task on learning from bad data and its connection with a data-dependent regularization.
- Photometric stereo for general isotropic reflectances by spherical linear interpolation11
This work presents a practical photometric stereo method that works with general isotropic reflectances and separately computes azimuth and elevation angles of the surface normal for each pixel by spherically interpolating the light source directions.
- Polarized 3D: synthesis of polarization and depth cues for enhanced 3D sensing11
A fusion of polarization and depth, a synergistic technique to exploit the high lateral resolution and compensate for the low depth resolution of traditional sensors is proposed.
- Color Correction and Compression for Multi-view Video Using H.264 Features11
By discarding most of the chrominance information before encoding and restoring them when decoding, this novel scheme can greatly improve the compression rate without much loss in visual quality, while at the same time produce color images with similar appearance to reference without correction.
- PolGS: Polarimetric Gaussian Splatting for Fast Reflective Surface Reconstruction10
The proposed PolGS is a arimetric arimetric aussian aussian latting model allowing fast reflective surface reconstruction in 10 minutes, and takes the shortest time while comparable shape reconstruction accuracy is measured by Chamfer Distance in millimeters.
- 10
- 10
- L-CAD: Language-based Colorization with Any-level Descriptions using Diffusion Priors10
A unified model to perform language-based colorization with any-level descriptions using the pretrained cross-modality generative model for its robust language understanding and rich color priors to handle the inherent ambiguity of any- level descriptions is proposed.
- 10
- 10
- AVI-Edit: Audio-sync Video Instance Editing with Granularity-Aware Mask Refiner9
A granularity-aware mask refiner that iteratively refines coarse user-provided masks into precise instance-level regions and a self-feedback audio agent to curate high-quality audio guidance, providing fine-grained temporal control is proposed.
- 9
- BokehDiff: Neural Lens Blur with One-Step Diffusion9
This work introduces BokehDiff, a novel lens blur rendering method that achieves physically accurate and visually appealing outcomes, with the help of generative diffusion prior, and proposes to synthesize photorealistic foregrounds with transparency with diffusion models, balancing authenticity and scene diversity.
- 9
- Self-Supervised Scene Dynamic Recovery from Rolling Shutter Images and Events9
An event-based SDR network within a self-supervised learning paradigm and an Event-based Inter/intra-frame Compensator (E-IC) is proposed to predict the per-pixel dynamic between arbitrary time intervals, including the temporal transition and spatial translation.
- 9
- Smooth Deep Image Generator from Noises9
A smooth generator is developed by investigating the tolerable input perturbation and integrating this smooth generator with a gradient penalized discriminator, and design smooth GAN that generates stable and high-quality images.
- 8
- PolarAnything: Diffusion-based Polarimetric Image Synthesis8
This work introduces a diffusion-based generative framework with an effective representation strategy that preserves the fidelity of polarization properties, capable of synthesizing polarization images from a single RGB input with both photorealism and physical accuracy.
- Polarization Guided Mask-Free Shadow Removal8
Pol-ShaRe is proposed, the first Polarization-guided image Shadow Removal solution, to remove shadow in a mask-free manner with fewer artifacts, and consists of a two-stage pipeline to relieve the ill-posedness and a neural network tailored to the pipeline to suppress the artifacts.
- 8
- Spin-UP: Spin Light for Natural Light Uncalibrated Photometric Stereo8
Spin-UP is proposed, an unsupervised method to tackle NaUPS in various environment lights and objects that uses a novel setup that captures the object's images on a rotatable platform, which mitigates NaUPS's ill-posedness by reducing unknowns and provides reliable priors to alleviate NaUPS's ambiguities.
- Non-Lambertian Multispectral Photometric Stereo via Spectral Reflectance Decomposition8
A spectral reflectance decomposition model to disentangle the spectral reflectances into a geometric component and a spectral component is presented and it is shown that the MPS problem for surfaces with a uniform material is equivalent to the conventional photometric stereo (CPS) with unknown light intensities.
- Conditional Image Repainting via Semantic Bridge and Piecewise Value Function8
This work improves the compositing by breaking through the latent ceiling using a novel piecewise value function and demonstrates on two datasets that the proposed techniques can better assist tackling conditional image repainting compared to the existing ones.
- Learning to remove reflections from windshield images8
A reflection removal framework specially designed for surveillance windshield images, with the guidance of depth of field features introduced from different levels of blurs, and constraint of specific semantic features for vehicle retrieval is proposed.
- 7
- Toward Deeper Emotional Reflection: Crafting Affective Image Filters With Generative Priors7
The Affective Image Filter (AIF) task, which aims to reflect visually-abstract emotions from text into visually-concrete images, thereby creating emotionally compelling results, is proposed.
- 7
- Spatio-Temporal Interactive Learning for Efficient Image Reconstruction of Spiking Cameras7
An efficient spatio-temporal interactive reconstruction network is proposed to jointly perform inter-frame feature alignment and intra-frame feature filtering in a coarse-to-fine manner and introduces a symmetric interactive attention block and a multi-motion field estimation block to further enhance the interaction capability of the overall network.
- 7
- 7
- 7
- 7
- AdsCVLR: Commercial Visual-Linguistic Representation Modeling in Sponsored Search7
This paper proposes a transformer architecture with Ads data on Commercial Visual-Linguistic Representation (AdsCVLR) with contrastive learning that naturally extends the transformer encoder with the complementary multi-modal inputs, serving as a strong aggregator of image-text features.
- Edge-preserving Near-light Photometric Stereo with Neural Surfaces7
A near-light photometric stereo method that faithfully preserves sharp depth edges in the 3D reconstruction and formulates the Lambertian albedo as a dependent variable resulting from the surface normal and depth, which is insusceptible to inaccurate depth initialization.
- 7
- Recent Progress in Shape from Polarization7
This chapter surveys recent progress in shape from polarization combined with different auxiliary information such as geometric cues, spectral cues, photometric cues, and deep learning, and introduces how polarization imaging benefits other vision tasks in addition to shape recovery.
- 7
- 7
- A self-calibrated photo-geometric depth camera7
This paper presents a practical 3D imaging system by combining the near-light photometric stereo and the speckle-based stereo matching method, which is compact in structure and suitable for multi-albedo targets.
- 7
- Convex Optimization Based Low-Rank Matrix Completion and Recovery for Photometric Stereo and Factor Classification7
This work presents a new approach to robustly solve the photometric stereo problem that uses an advanced convex optimization technique that is guaranteed to find the correct low-rank matrix by simultaneously fixing its missing and erroneous entries.
- 6
- 6
- Color4E: Event Demosaicing for Full-color Event Guided Image Deblurring6
A novel network called Color4E is proposed to enhance the color restoration quality for the image deblurring task and enables high-quality event-based image deblurring compared to state-of-the-art methods.
- 6
- 6
- 6
- When Visible Light (Backscatter) Communication Meets Neuromorphic Cameras in V2X6
This paper proposes NeuromorphicVLC, a visible light communication system equipped with neuromorphic cameras as optical receivers to improve its performance, and develops a complete signal processing pipeline to detect the VLC transmitters and demodulate the messages.
- 6
- 6
- 6
- 6
- 5
- No Redundancy, No Stall: Lightweight Streaming 3D Gaussian Splatting for Real-time Rendering5
LS-Gaussian, an algorithm/hardware co-design framework for lightweight streaming 3D rendering that predicts the workload for each tile based on viewpoint transformation to enable more balanced parallel computation and co-designs a customized 3DGS accelerator to support the workload-aware mapping in real-time.
- Self-supervised Shutter Unrolling with Events5
An event-based CGVR network within a self-supervised learning paradigm is proposed, and an Event-based Inter/intra-frame Compensator (E-IC) is proposed to predict the per-pixel dynamic between arbitrary time intervals, including the temporal transition and spatial translation.
- 5
- 5
- 5
- 5
- 5
- 5
- Group Contextual Encoding for 3D Point Clouds5
A group contextual encoding method is proposed, which divides the channel into groups and then performs encoding on group-divided feature vectors, which facilitates learning of global context in grouped subspace for 3D point clouds.
- 5
- 5
- 5
- EvTurb: Event Camera Guided Turbulence Removal4
EvTurb is an event guided turbulence removal framework that leverages high-speed event streams to decouple blur and tilt effects, and presents TurbEvent, the first real-captured dataset featuring diverse turbulence scenarios.
- 4
- 4
- VIRES: Video Instance Repainting via Sketch and Text Guided Generation4
This work introduces VIRES, a video instance repainting method with sketch and text guidance, enabling video instance repainting, replacement, generation, and removal, and proposes the Sequential ControlNet with the standardized self-scaling, which effectively extracts structure layouts and adaptively captures high-contrast sketch details.
- PlaNet: Learning to Mitigate Atmospheric Turbulence in Planetary Images4
A vertical distance-aware turbulence simulation pipeline is proposed to synthesize realistic planetary images in accordance with their unique degeneration patterns at a large scale with affordable computational cost, and a neural network is designed to mitigate the turbulence with flexible input frames by adopting an edge-based supervision strategy to handle the background scarcity issue.
- 4
- Slow and Weak Attractor Computation Embedded in Fast and Strong E-I Balanced Neural Dynamics4
This study investigates the compatibility of continuous attractor neural networks (CANNs) and E-INNs and finds that a neural circuit can exhibit both the traits of CANNs and E-INNs when the neuronal synapses consist of two sets: one set is strong and fast for irregular firing, and the other set is weak and slow for attractor dynamics.
- 4
- 4
- 4
- 4
- 4
- CDIR: LoRA-Inspired Attention for Efficient Composite Degradation Image Restoration3
The proposed network, CDIR, achieves state-of-the-art performance with significantly reduced complexity and fast inference speed, and shows strong adaptability to various task-specific image restoration scenarios, such as dehazing, desnowing, and deraining.
- AdaptiveAE: An Adaptive Exposure Strategy for HDR Capturing in Dynamic Scenes3
AdaptiveAE is a reinforcement learning-based method that optimizes the selection of shutter speed and ISO combinations to maximize HDR reconstruction quality in dynamic environments and can adaptively select optimal ISO and shutter speed sequences based on a user-defined exposure time budget.
- 3
- 3
- 3
- 3
- Quality-Improved and Property-Preserved Polarimetric Imaging via Complementarily Fusing3
This work proposes a polarimetric imaging framework that can produce clean and clear polarized snapshots by complementarily fusing a degraded pair of noisy and blurry ones by adopting a neural network-based three-phase fusing scheme with specially-designed modules tailored to each phase.
- 3
- Instance Contour Adjustment via Structure-Driven CNN3
The meaningfulness of image textures with clear and user-desired contours are guaranteed by the structure-driven CNN, which does not require any semantic label as input, which ensures its well generalization capa-bility.
- See Through the Windshield from Surveillance Camera3
A windshield image restoration generative adversarial network (WIRE-GAN) to restore and enhance the visibility of windshield images and introduces a novel content-preserving strategy into the proposed weakly-supervised framework to generate more semantically consistent results.
- 3
- 2
- PacTure: Efficient PBR Texture Generation on Packed Views with Visual Autoregressive Models2
View packing is introduced, a novel technique that significantly increases the effective resolution for each view during multi-view generation, without imposing additional inference cost, to reduce the inferencing cost and enable fine-grained control and multi-domain generation within the next-scale prediction autoregressive framework.
- 2
- 2
- Personalized Image Filter: Mastering Your Photographic Style2
A Personalized Image Filter (PIF) is proposed, based on a pretrained text-to-image diffusion model, which enables PIF to learn the average appearance of photographic concepts, as well as how to adjust them according to text prompts, by optimizing the prompts for the photographic concepts.
- 2
- 2
- 2
- 2
- Technique Report of CVPR 2024 PBDL Challenges2
This technical report summarizes the outcomes of the Physics-Based Vision Meets Deep Learning (PBDL) 2024 challenge, held in CVPR 2024 workshop, focusing on Low-Light Enhancement and Detection as well as High Dynamic Range (HDR) Imaging.
- Surface Geometry Processing: An Efficient Normal-Based Detail Representation2
An efficient surface detail processing framework in 2D normal domain is introduced, which extracts new normal feature representations as the carrier of micro geometry structures that are illustrated both theoretically and empirically in this article.
- 1
- 1
- 1
- ReContraster: Making Your Posters Stand Out with Regional Contrast1
ReContraster is proposed, the first training-free model to leverage regional contrast to make posters stand out, and introduces the compositional multi-agent system to identify elements, organize layout, and evaluate generated poster candidates.
- Asynchronous Event Error-Minimizing Noise for Safeguarding Event Dataset1
This work proposes the first unlearnable event stream generation method to prevent unauthorized training from event datasets, and proposes a new form of asynchronous event error-minimizing noise to perturb event streams, tricking the unauthorized model into learning embedded noise instead of realistic features.
- EDeF-Net: Spatio-temporal Association Network for Flicker Removal in Event Streams1
EDeF-Net is proposed, an Event streams DeFlickering Network that effectively leverages the spatio-temporal correlation of event streams by modeling both the inter-channel temporal attention and inter-patch spatial attention and yields performance improvements on down-stream applications such as event-based optical flow estimation and object tracking.
- 1
- DreamPolish: Domain Score Distillation With Progressive Geometry Generation1
This work introduces DreamPolish, a text-to-3D generation model that excels in producing refined geometry and high-quality textures and proposes to add a surface polishing stage with only a few training steps, which can effectively refine the artifacts attributed to limited guidance from previous stages and produce 3D objects with more desirable geometry.
- Spike Stream Denoising via Spike Camera Simulation1
A systematic noise model for spike camera based on its unique circuit is proposed and a tailored spike stream denoising framework (DnSS) where denoised spike stream is obtained by decoding inferred inter-spike intervals is designed.
- EvPlug: Learn a Plug-and-Play Module for Event and Image Fusion1
EvPlug is proposed, which learns a plug-and-play event and image fusion module from the supervision of the existing RGB-based model, endowing the RGB-based model to be robust to HDR and fast motion scenes while enabling high temporal resolution inference.
- NeuralMPS: Non-Lambertian Multispectral Photometric Stereo via Spectral Reflectance Decomposition1
A deep neural network named NeuralMPS is proposed to solve the MPS problem under general non-Lambertian spectral reflectance decomposition (SRD) model to disentangle the spectral re-partialance into geometric components and spectral components and shows that the Mps problem for surfaces with a uniform material is equivalent to the conventional photometric stereo (CPS) with unknown light intensities.
- NeIF: Representing General Reflectance as Neural Intrinsics Fields for Uncalibrated Photometric Stereo1
The key idea is to represent the reflectance as four neural intrinsics fields, i.e., position, light, specular, and shadow, based on which the neural light field is implicitly associated with light clues of specular reflectance and cast shadow, and solves UPS in an unsupervised manner.
- 1
- 1
- Spectral Representation via Data-Guided Sparsity for Hyperspectral Image Super-Resolution1
Since the proposed method adaptively adjusts the sparsity in the spectral representation based on the local content of the available high resolution RGB image, it can produce more robust spectral representation for recovering the target high resolution hyperspectral image.
- 1
- Snapshot On-chip HDR ROIC Architectures1
Novel digital readout integrated circuits (DROICs) are described that achieve snapshot on-chip high dynamic range imaging where most commercial systems require a multiple exposure acquisition.
- –
- –
- RefMover: Diffusion-Based Single Image Reflection Removal With Language and Region Guidance–
RefMover is introduced, a multi-instruction diffusion framework for SIRR that bridges the semantic-spatial gap by combining language and region guidance and features a pipeline with a streamlined conditional architecture that mitigates condition conflicts and preserves details.
- –
- –
- –
- –
- –
- Large Kernel Network for Efficient Image Restoration–
This paper proposes a simple and efficient large kernel network for image restoration, termed ArtIR, motivated by the channel redundancy observed in image restoration models, and applies adaptive large-kernel operations to a collapsed single channel and employs an ultra-lightweight channel attention mechanism to restore channel diversity.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Dense Metric Depth Estimation via Event-based Differential Focus Volume Prompting–
This paper builds Event-based Differential Focus Volumes (EDFV) using events triggered through focus sweeping, which are subsequently transformed into sparse metric depth maps and utilized for prompting dense depth estimation via the proposed Event-based Depth Prompting Network.
- SymmeTac: Symmetric Color LED Driven Efficient Photometric Stereo Reconstruction Methods for Camera-based Tactile Sensors–
This work designs orthogonal red and blue LEDs as illumination to acquire four observation maps using channel-splitting in a two-shot manner, and develops a two-shot photometric stereo theory, which can estimate accurate surface normal and greatly reduce the computing overhead in magnitude.
- Shape and Albedo Recovery by Your Phone using Stereoscopic Flash and No-Flash Photography–
This paper presents a high-fidelity shape and albedo recovery method that only requires a stereo camera and a flashlight, a typical camera setup equipped in many off-the-shelf smartphones.
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.