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
- Marc PollefeysSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works114 from public data
- Video Object Segmentation Using Space-Time Memory Networks903
This work proposes a novel solution for semi-supervised video object segmentation by leveraging memory networks and learning to read relevant information from all available sources to better handle the challenges such as appearance changes and occlussions.
- 610
- 423
- 363
- EPINET: A Fully-Convolutional Neural Network Using Epipolar Geometry for Depth from Light Field Images319
This paper introduces a fast and accurate light field depth estimation method based on a fully-convolutional neural network that achieved the top rank in the HCI 4D Light Field Benchmark on most metrics, and demonstrates the effectiveness of the proposed method on real-world light-field images.
- 284
- 268
- 171
- Copy-and-Paste Networks for Deep Video Inpainting146
A novel DNN-based framework called the Copy-and-Paste Networks for video inpainting that takes advantage of additional information in other frames of the video to significantly improve the lane detection accuracy on road videos.
- Visual enhancement of old documents with hyperspectral imaging129
This work demonstrates how to use the invisible bands to improve the visual quality of text-based documents corrupted with undesired artifacts such as ink-bleed, ink-corrosion, and foxing.
- Onion-Peel Networks for Deep Video Completion116
An asymmetric attention block is proposed that computes similarities between the hole boundary pixels in the target and the non-hole pixel in the references in a non-local manner and can have an unlimited spatial-temporal window size and fill the holes with globally coherent contents.
- 115
- Approaching the computational color constancy as a classification problem through deep learning109
A deep learning framework for the illumination estimation problem is adopted and the convolutional neural network is trained to solve the problem by casting the color constancy problem as an illumination classification problem.
- Building Emotional Machines: Recognizing Image Emotions Through Deep Neural Networks97
This paper focuses on two high-level features, the object and the background, and assumes that the semantic information in images is a good cue for predicting emotions, and builds an emotion-based feedforward deep neural network that produces the emotion values of a given image.
- 93
- 92
- 86
- 85
- A Generalized Framework for Video Instance Segmentation83
A Generalized framework for VIS, namely GenVIS, that achieves state-of-the-art performance on challenging benchmarks without designing complicated architectures or requiring extra post-processing, and introduces a memory that effectively acquires information from previous states.
- 81
- Fast User-Guided Video Object Segmentation by Interaction-And-Propagation Networks70
This work proposes a new multi-round training scheme for the interactive video object segmentation so that the networks can learn how to understand the user's intention and update incorrect estimations during the training.
- 61
- 60
- 59
- 58
- 58
- NTIRE 2020 Challenge on Perceptual Extreme Super-Resolution: Methods and Results54
This paper reviews the NTIRE 2020 challenge on perceptual extreme super-resolution with focus on proposed solutions and results and gauges the state-of-the-art in single image super- resolution.
- 53
- UniSkill: Imitating Human Videos via Cross-Embodiment Skill Representations47
UniSkill is a novel framework that learns embodiment-agnostic skill representations from large-scale cross-embodiment video data without any labels, enabling skills extracted from human video prompts to effectively transfer to robot policies trained only on robot data.
- 46
- 43
- 43
- 43
- UBoCo: Unsupervised Boundary Contrastive Learning for Generic Event Boundary Detection40
A novel framework for unsupervised/supervised GEBD, by using the Temporal Self-similarity Matrix (TSM) as the video representation, and the new Recursive TSM Parsing (RTP) algorithm exploits local diagonal patterns in TSM to detect boundaries, and it is combined with the Boundary Contrastive (BoCo) loss to train the encoder to generate more informative TSMs.
- 40
- VISOLO: Grid-Based Space-Time Aggregation for Efficient Online Video Instance Segmentation38
The design fully takes advantage of previous information in a grid form for all tasks in VIS in an efficient way, and the new state-of-the-art accuracy and speed on YouTube-VIS 2019 and 2021 datasets among online VIS methods are achieved.
- Multi-Granular Spatio-Temporal Token Merging for Training-Free Acceleration of Video LLMs37
A training-free spatio-temporal token merging method, named STTM, to exploit local spatial and temporal redundancy in video data which has been overlooked in prior work, and outperforms existing token reduction methods across six video QA benchmarks.
- Temporally smooth online action detection using cycle-consistent future anticipation37
This paper presents a novel solution for online action detection by using a simple yet effective RNN-based networks called the Future Anticipation and Temporally Smoothing network (FATSnet), consisting of a module for anticipating the future that can be trained in an unsupervised manner with the cycle-consistency loss.
- 36
- 33
- Latent Space Super-Resolution for Higher-Resolution Image Generation with Diffusion Models32
This paper proposes LSRNA, a novel framework for higher-resolution image generation using diffusion models by leveraging super-resolution directly in the latent space by combining Latent space Super-Resolution for manifold alignment and Region-wise Noise Addition (RNA) to enhance high-frequency details.
- Single‐shot High Dynamic Range Imaging Using Coded Electronic Shutter31
This paper proposes an HDR imaging approach using the coded electronic shutter which can capture a scene with row‐wise varying exposures in a single image, and enables a direct extension of the dynamic range of the captured image without using multiple images, by photometrically calibrating rows with different exposures.
- 28
- 27
- Modelling the Scene Dependent Imaging in Cameras with a Deep Neural Network27
A datadriven approach to learn the scene dependent and locally varying image processing inside cameras under the automode that incorporates both the global and the local scene context into pixel-wise features via multi-scale pyramid of learnable histogram layers.
- 27
- Shepherding Slots to Objects: Towards Stable and Robust Object-Centric Learning24
A novel OCL framework for single-view images, SLot Attention via SHepherding (SLASH), which consists of two simple- yet-effective modules on top of Slot Attention, and a weak semi-supervision approach for OCL that can be used without any assistant annotation during the inference.
- 24
- Omni-RGPT: Unifying Image and Video Region-level Understanding via Token Marks23
Token Mark, a set of tokens highlighting the target regions within the visual feature space that are directly embedded into spatial regions using region prompts and simultaneously incorporated into the text prompt to specify the target, establishes a direct connection between visual and text tokens.
- Attentive Illumination Decomposition Model for Multi-Illuminant White Balancing21
A deep white balancing model that leverages the slot attention, where each slot is in charge of representing individual illuminants, which achieves the state-of-the-art performance on both single- and multi-illuminant WB benchmarks, and also offers additional information such as the number of illuminants in the scene and their chromaticity.
- Deep Space-Time Video Upsampling Networks21
An end-to-end DNN framework for the space-time video upsampling by efficiently merging VSR and FI into a joint framework is proposed and a novel weighting scheme is proposed to fuse input frames effectively without explicit motion compensation for efficient processing of videos.
- 20
- Nonuniform Lattice Regression for Modeling the Camera Imaging Pipeline20
This work describes a method to construct a sparse lookup table (LUT) that is effective in modeling the camera imaging pipeline that maps a RAW camera values to their sRGB output and adapts the LUT lattice to better fit the 3D gamut-mapping function.
- Generating Fluttering Patterns with Low Autocorrelation for Coded Exposure Imaging19
It is shown that the concept of low autocorrelation binary sequences, which has been successfully exploited in the field of information theory, can be applied to generate shutter fluttering patterns and a new measure for good binary sequences is proposed.
- 18
- Probabilistic moving least squares with spatial constraints for nonlinear color transfer between images18
A novel color transfer framework based on a scattered point interpolation scheme that solves for a full nonlinear and nonparametric color mapping in the 3D RGB color space by employing the moving least squares framework is introduced.
- 17
- High-Resolution Multiscale Panoramic Mosaics from Pan-Tilt-Zoom Cameras.17
This paper presents an automatic software based approach for building extremely high resolution panoramic mosaics from images captured by an off-the-shelf pan tilt zoom camera, and shows results from two datasets captured from cameras placed in an uncontrolled outdoor scene.
- A Sliding Window Scheme for Online Temporal Action Localization16
Online Anchor Transformer (OAT) is proposed to extend the anchor-based action localization model to the online setting and introduces an online-applicable post-processing method that suppresses repetitive action proposals.
- Joint radiometric calibration and feature tracking system with an application to stereo15
A system for tracking features and estimating the radiometric response function of the camera and the exposure difference between frames simultaneously is introduced and applied to perform structure-from-motion and stereo to reconstruct a texture-mapped 3D surface from a video taken in a high dynamic range environment.
- ORIDa: Object-centric Real-world Image Composition Dataset13
This work introduces ORIDa (Object-centric Real-world Image Composition Dataset), a large-scale, real-captured dataset containing over 30,000 images featuring 200 unique objects, each of which is presented across varied positions and scenes and is the first publicly available dataset with its scale and complexity for real-world image composition.
- Single-shot Path Integrated Panoptic Segmentation13
This paper proposes to resolve panoptic segmentation in single-shot by integrating the execution flows, and generates a unified feature map called Panoptic-Feature, which includes the information of both things and stuffs.
- Cross-Identity Motion Transfer for Arbitrary Objects Through Pose-Attentive Video Reassembling13
Experimental results validate that the proposed attention-based networks for transferring motions between arbitrary objects produces visually pleasing results in various object domains, showing better performances compared to previous works.
- CCMNet: Leveraging Calibrated Color Correction Matrices for Cross-Camera Color Constancy12
A learning-based method for crosscamera color constancy that generalizes to new cameras without retraining and achieves state-of-the-art cross-camera color constancy while remaining lightweight and relying only on data readily available in camera ISPs is introduced.
- ATTIQA: Generalizable Image Quality Feature Extractor Using Attribute-Aware Pretraining12
A novel pretraining framework is proposed that constructs a generalizable representation for IQA by selectively extracting quality-related knowledge from VLM and leveraging the scalability of large datasets, achieving state-of-the-art performance on multiple IQA datasets and exhibits remarkable generalization capabilities.
- 12
- 12
- Effects of Path-Finding Algorithms on the Labeling of the Centerlines of Circle of Willis Arteries11
Three-dimensional time-of-flight magnetic resonance angiography (MRA) images from the publicly available dataset were considered for this study and the Dijkstra and A* algorithms showed similar accuracy in path-finding, and they were comparable in the speed in the circle of Willis arterial segments.
- Teaching Machines to Understand Baseball Games: Large-Scale Baseball Video Database for Multiple Video Understanding Tasks10
A new large-scale baseball video dataset called the BBDB, which is produced semi-automatically by using play-by-play texts available online, and can be used for many video understanding tasks including video recognition, localization, text-video alignment, video highlight generation, and data imbalance problem.
- Brush-and-drag10
An effective and easy-to-use brush-and-drag interface that allows the user to interactively explore and compare photos within a broader scene context and demonstrates to offer effective interactions in a user study.
- Soft-Landing Strategy for Alleviating the Task Discrepancy Problem in Temporal Action Localization Tasks9
This work introduces Soft-Landing (SoLa) strategy, an efficient yet effective framework to bridge the transferability gap between the pretrained encoder and the downstream tasks by incorporating a light-weight neural network on top of the frozen encoder.
- Colorization for Single Image Super Resolution9
A procedure based on image colorization and back-projection to perform color assignment guided by the super-resolution luminance channel is outlined and it is found that this procedure produces better results both quantitatively and qualitatively than existing approaches.
- Seam360GS: Seamless 360° Gaussian Splatting from Real-World Omnidirectional Images8
A novel calibration framework is introduced that incorporates a dual-fisheye camera model into the 3D Gaussian splatting pipeline and transforms imperfect omnidirectional inputs into flawless novel view synthesis.
- Exploring Scalability of Self-Training for Open-Vocabulary Temporal Action Localization7
This paper explores the scalability of self-training with unlabeled YouTube videos for OV-TAL and identifies limitations in existing OV-TAL evaluation schemes and proposes a new benchmark for thorough assessment.
- ExploreGS: Explorable 3D Scene Reconstruction with Virtual Camera Samplings and Diffusion Priors7
This work proposes a 3DGS-based pipeline that generates additional training views to enhance reconstruction, and introduces an information-gain-driven virtual camera placement strategy to maximize scene coverage, followed by video diffusion priors to refine rendered results.
- Representing 3D Shapes with 64 Latent Vectors for 3D Diffusion Models6
This paper introduces COD-VAE that encodes 3D shapes into a COmpact set of 1D latent vectors without sacrificing quality and proposes uncertainty-guided token pruning, which allocates resources adaptively by skipping computations in simpler regions and improves the decoder efficiency.
- 6
- 6
- 6
- 5
- An Improved Path-Finding Method for the Tracking of Centerlines of Tortuous Internal Carotid Arteries in MR Angiography5
The proposed method has been shown to be more suitable as a path-finding procedure than the existing methods, particularly in cases where there is more than one centerline resulting from over-segmentation of a highly tortuous artery.
- Cannot See the Forest for the Trees: Aggregating Multiple Viewpoints to Better Classify Objects in Videos5
A set classifier that improves accuracy of classifying tracklets by aggregating information from multiple viewpoints contained in a tracklet is presented, which is plug-and-playable to existing object trackers, and highly improves the performance of long-tailed object tracking.
- 5
- Robust Stereo Matching Using Probabilistic Laplacian Surface Propagation5
Experimental results show that the PLSP outperforms state-of-the-art robust cost function based methods and other propagation methods for the stereo matching under radiometric variations.
- Color-Aware Regularization for Gradient Domain Image Manipulation5
This work proposes a color-aware regularization for use with gradient domain image manipulation to avoid color shift artifacts and uses an anisotropic Mahalanobis distance to control output colors to better fit original distributions.
- 5
- Polygonal Point Set Tracking4
This paper proposes a new method to track the corresponding points between frames by the global-local alignment with delicately designed losses and regularization terms and introduces a novel learning strategy using synthetic and VOS datasets that makes it possible to tackle the problem without developing the point correspondence dataset.
- Rethinking State Tracking in Recurrent Models Through Error Control Dynamics3
It is proved that affine recurrent networks, a class of models encompassing State-Space Models and Linear Attention, cannot correct errors along state-separating subspaces once they preserve state representations, and established that robust state tracking is determined not only by an architecture's theoretical expressivity but crucially by its error control.
- Where Did It Go Wrong? Process-Level Evaluation of Web Agents with Semantic State Tracking3
This work conducts a process-level analysis of web agents, providing fine-grained and actionable insight into where and how each agent should be improved, and introduces WebStep, a benchmark of 1,800 task instances with controlled difficulty and automatic semantic state tracking.
- 3
- 3
- Robust and Consistent Online Video Instance Segmentation via Instance Mask Propagation3
RoCoVIS, a simple yet effective approach that integrates segmentation and tracking to provide consistent online VIS, is proposed, an end-to-end sequential learning where object queries are propagated through mask predictions, improving the accuracy of temporal instance mapping at the pixel level.
- IF-MDM: Implicit Face Motion Diffusion Model for High-Fidelity Realtime Talking Head Generation3
Although implicit motion lacks the spatial disentanglement of explicit models, which complicates alignment with subtle lip movements, this model provides motion controllability to optimize the trade-off between motion intensity and visual quality during inference.
- Radiometric calibration methods from image sequences3
This dissertation presents several algorithms to compute the radiometric properties of a camera which enable us to find the relationship between the image brightness and the scene radiance and proposes a radiometric calibration method suited for a set of images of an outdoor scene taken at a regular interval over a period of time.
- Decomposed Attention Fusion in MLLMs for Training-Free Video Reasoning Segmentation2
Decomposed Attention Fusion (DecAF), which refines these maps through two mechanisms: contrastive object-background fusion and complementary video-frame fusion, outperforms training-free methods and achieves performance comparable to training-based methods on both referring and reasoning VOS benchmarks.
- 2
- 2
- 2
- 1
- Object Aware Egocentric Online Action Detection1
An Object-Aware Module is introduced that integrates egocentric-specific priors into existing OAD frameworks, enhancing first-person footage interpretation and improves scene understanding in detecting actions.
- 1
- –
- –
- –
- Unsupervised Monocular 3D Keypoint Discovery from Multi-View Diffusion Priors–
This paper introduces KeyDiff3D, a framework that can accurately predict 3D keypoints from a single image, thus eliminating the need for such expensive data acquisitions and introducing a pipeline that enables manipulation of 3D objects generated by the diffusion model.
- –
- Leveraging Image Augmentation for Object Manipulation: Towards Interpretable Controllability in Object-Centric Learning–
A novel method is introduced, Slot 006 Attention with Image Augmentation (SlotAug), to explore the possibility of learning interpretable controllability over slots in a self-supervised manner by utilizing image aug-009 mentation.
- –
- –
- –
- –
- Creating Picture Legends for Group Photos–
An efficient method for producing picture legends for group photos that combines face detection with human shape priors into an interactive selection framework to allow users to quickly segment the individuals in a group photo is introduced.
- –
- –
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 2026-10-14. 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.