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
- Shaogang GongSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works187 from public data
- DehazeNet: An End-to-End System for Single Image Haze Removal3,391
This paper proposes a trainable end-to-end system called DehazeNet, for medium transmission estimation, which takes a hazy image as input, and outputs its medium transmission map that is subsequently used to recover a haze-free image via atmospheric scattering model.
- Fantasia3D: Disentangling Geometry and Appearance for High-quality Text-to-3D Content Creation843
Key to Fantasia3D is the disentangled modeling and learning of geometry and appearance, which is more compatible with popular graphics engines, supporting relighting, editing, and physical simulation of the generated 3D assets.
- Human Action Recognition Using Factorized Spatio-Temporal Convolutional Networks553
Factorized spatio-temporal convolutional networks (FstCN) are proposed that factorize the original 3D convolution kernel learning as a sequential process of learning 2D spatial kernels in the lower layers, followed by learning 1D temporal kernel in the upper layers.
- Frustum ConvNet: Sliding Frustums to Aggregate Local Point-Wise Features for Amodal 3D Object Detection511
A novel method termed Frustum ConvNet (F-ConvNet), which aggregates point-wise features as frustum-level feature vectors, and arrays these feature vectors as a feature map for use of its subsequent component of fully convolutional network (FCN).
- Frustum ConvNet: Sliding Frustums to Aggregate Local Point-Wise Features for Amodal511
A novel method termed Frustum ConvNet (F-ConvNet), which aggregates point-wise features as frustum-level feature vectors, and arrays these feature vectors as a feature map for use of its subsequent component of fully convolutional network (FCN).
- Domain-Symmetric Networks for Adversarial Domain Adaptation356
This paper proposes a new domain adaptation method called Domain-Symmetric Networks (SymNets), which is based on a symmetric design of source and target task classifiers, based on which an additional classifier is constructed that shares with them its layer neurons.
- Unsupervised Domain Adaptation via Structurally Regularized Deep Clustering331
This work describes the proposed method as Structurally Regularized Deep Clustering (SRDC), where it enhances target discrimination with clustering of intermediate network features, and enhance structural regularization with soft selection of less divergent source examples.
- 308
- GS-IR: 3D Gaussian Splatting for Inverse Rendering300
The GS-IR is a novel inverse rendering approach based on 3D Gaussian Splatting that leverages forward mapping volume rendering to achieve photorealistic novel view synthesis and relighting results and demonstrates the superiority of the method over baseline methods through qualitative and quantitative evaluations of various challenging scenes.
- 293
- Exact Feature Distribution Matching for Arbitrary Style Transfer and Domain Generalization269
This work proposes to perform Exact Feature Distribution Matching (EFDM) by exactly matching the empirical Cumulative Distribution Functions (eCDFs) of image features, which could be implemented by applying the Exact Histogram Matching(EHM) in the image feature space.
- Discriminative Adversarial Domain Adaptation234
Discriminative Adversarial Domain Adaptation has a novel adversarial objective that encourages a mutually inhibitory relation between category and domain predictions for any input instance and it defines a minimax game that can promote the joint distribution alignment.
- Deep Mesh Reconstruction From Single RGB Images via Topology Modification Networks225
This paper presents an end-to-end single-view mesh reconstruction framework that is able to generate high-quality meshes with complex topologies from a single genus-0 template mesh and outperforms the current state-of-the-art methods both qualitatively and quantitatively.
- 215
- 215
- Lattice Long Short-Term Memory for Human Action Recognition209
This work proposes Lattice-LSTM (L2STM), which extends LSTM by learning independent hidden state transitions of memory cells for individual spatial locations, which effectively enhances the ability to model dynamics across time and addresses the non-stationary issue of long-term motion dynamics without significantly increasing the model complexity.
- Object as Hotspots: An Anchor-Free 3D Object Detection Approach via Firing of Hotspots207
Inspired by compositional models, an object is represented as composition of its interior non-empty voxels, termed hotspots, and the spatial relations of hotspots gives rise to the representation of Object as Hotspots (OHS).
- SAM-6D: Segment Anything Model Meets Zero-Shot 6D Object Pose Estimation204
SAM-6D, a novel framework designed to realize the task of zero-shot 6D object pose estimation through two steps, including instance segmentation and pose estimation of novel objects, outperforms the existing methods on the seven core datasets of the BOP Benchmark.
- 3D AffordanceNet: A Benchmark for Visual Object Affordance Understanding194
Comprehensive results on the contributed dataset show the promise of visual affordance understanding as a valuable yet challenging benchmark.
- DualPoseNet: Category-level 6D Object Pose and Size Estimation Using Dual Pose Network with Refined Learning of Pose Consistency190
A new method of Dual Pose Network with refined learning of pose consistency for category-level 6D object pose and size estimation, shortened as DualPoseNet, which outperforms existing methods with a large margin in the regime of high precision.
- Towards Understanding Cross and Self-Attention in Stable Diffusion for Text-Guided Image Editing188
An in-depth probing analysis is conducted and it is demonstrated that cross-attention maps in Stable Diffusion often contain object attribution information, which can result in editing failures, and proposes a simplified, yet more stable and efficient, tuning-free procedure that modifies only the self-attention maps of specified attention layers during the denoising process.
- 184
- Instance Segmentation in 3D Scenes using Semantic Superpoint Tree Networks177
An end-to-end solution of Semantic Superpoint Tree Network (SSTNet) for proposing object instances from scene points, and design in SSTNet a refinement module, termed CliqueNet, to prune superpoints that may be wrongly grouped into instance proposals.
- Orthogonal Deep Neural Networks169
This paper proves that DNNs are of local isometry on data distributions of practical interest, and establishes a new generalization error bound that is both scale- and range-sensitive to singular value spectrum of each of networks’ weight matrices.
- Surface Reconstruction From Point Clouds: A Survey and a Benchmark166
The present paper contributes a large-scale benchmarking dataset consisting of both synthetic and real-scanned data that takes into account various sensing imperfections that are commonly encountered in practical depth scanning, and conducts thorough empirical studies by comparing existing methods on the constructed benchmark.
- 162
- 155
- 150
- EPMF: Efficient Perception-Aware Multi-Sensor Fusion for 3D Semantic Segmentation139
This work investigates a collaborative fusion scheme called perception-aware multi-sensor fusion (PMF) to effectively exploit perceptual information from two modalities, namely, appearance information from RGB images and spatio-depth information from point clouds, and proposes an improved version of PMF, i.e., EPMF, which is more efficient and effective.
- 139
- HEMlets Pose: Learning Part-Centric Heatmap Triplets for Accurate 3D Human Pose Estimation138
This work attempts to address the uncertainty of lifting the detected 2D joints to the 3D space by introducing an intermediate state - Part-Centric Heatmap Triplets (HEMlets), which shortens the gap between the 2D observation and the3D interpretation.
- 130
- 130
- PARN: Position-Aware Relation Networks for Few-Shot Learning126
This paper proposes a position-aware relation network (PARN) to learn a more flexible and robust metric ability for few-shot learning, and introduces a deformable feature extractor (DFE) to extract more efficient features and design a dual correlation attention mechanism (DCA) to deal with its inherent local connectivity.
- Robust and Practical Face Recognition via Structured Sparsity125
This paper introduces a class of structured sparsity-inducing norms into the SRC framework, to model various corruptions in face images caused by misalignment, shadow, illumination variation, and occlusion, and develops an automatic face alignment method based on minimizing the structuredSparsity norm.
- 121
- Fine-Grained Visual Categorization Using Meta-learning Optimization with Sample Selection of Auxiliary Data115
A new deep FGVC model termed MetaFGNet is proposed, based on a novel regularized meta-learning objective, which aims to guide the learning of network parameters so that they are optimal for adapting to the target FGVC task.
- 115
- MTL-NAS: Task-Agnostic Neural Architecture Search Towards General-Purpose Multi-Task Learning113
A novel single-shot gradient-based search algorithm that closes the performance gap between the searched architectures and the final evaluation architecture and achieves a single model, which can be directly used for evaluation without (re-)training from scratch.
- TANGO: Text-driven Photorealistic and Robust 3D Stylization via Lighting Decomposition112
TANGO enables photorealistic 3D style transfer by automatically predicting reflectance effects even for bare, low-quality meshes, without training on a task-specific dataset, and outperforms existing methods of text-driven 3D Style transfer in terms of photorealism quality, consistency of 3D geometry, and robustness when stylizing low- quality meshes.
- JOLO-GCN: Mining Joint-Centered Light-Weight Information for Skeleton-Based Action Recognition110
This paper proposes a novel framework for employing human pose skeleton and joint-centered light-weight information jointly in a two-stream graph convolutional network, namely, JOLO-GCN, and uses Joint-aligned optical Flow Patches to capture the local subtle motion around each joint as the pivotal joint- centered visual information.
- Unsupervised Multi-Class Domain Adaptation: Theory, Algorithms, and Practice101
An algorithmic framework of Multi-class Domain-adversarial learning Networks (McDalNets) is developed, and its different instantiations via surrogate learning objectives either coincide with or resemble a few recently popular methods, thus (partially) underscoring their practical effectiveness.
- VISTA: Boosting 3D Object Detection via Dual Cross-VIew SpaTial Attention97
This paper proposes to adaptively fuse multi-view features in a global spatial context via Dual Cross- VIew SpaTial Attention (VISTA), a novel plug-and-play fusion module wherein the multi-layer perceptron widely adopted in standard attention modules is replaced with a convolutional one.
- Improving Training of Deep Neural Networks via Singular Value Bounding97
This work proposes to constrain the solutions of weight matrices in the orthogonal feasible set during the whole process of network training, and achieves this by a simple yet effective method called Singular Value Bounding (SVB).
- A Skeleton-Bridged Deep Learning Approach for Generating Meshes of Complex Topologies From Single RGB Images92
Qualitative and quantitative results on representative object categories of both simple and complex topologies demonstrate the superiority of the proposed skeleton-bridged, stage-wise learning approach over existing methods.
- 90
- SA-ConvONet: Sign-Agnostic Optimization of Convolutional Occupancy Networks86
This work proposes to learn implicit surface reconstruction by sign-agnostic optimization of convolutional occupancy networks, to simultaneously achieve advanced scalability to large-scale scenes, generality to novel shapes, and applicability to raw scans in a unified framework.
- 83
- Cascading Convolutional Color Constancy82
This paper introduces a novel algorithm – Cascading Convolutional Color Constancy (in short, C4) to improve robustness of regression learning and achieve stable generalization capability across datasets (different cameras and scenes) in a unique framework.
- Label Propagation with Augmented Anchors: A Simple Semi-supervised Learning Baseline for Unsupervised Domain Adaptation78
This work suggests a new algorithm of Label Propagation with Augmented Anchors (A$^2$LP), which could potentially improve LP via generation of unlabeled virtual instances (i.e., the augmented anchors) with high-confidence label predictions and tackle the domain-shift challenge of UDA by alternating between pseudo labeling via A$^1$LP and domain-invariant feature learning.
- VI-Net: Boosting Category-level 6D Object Pose Estimation via Learning Decoupled Rotations on the Spherical Representations76
A novel rotation estimation network is proposed, termed as VI-Net, to make the task easier by decoupling the rotation as the combination of a viewpoint rotation and an in-plane rotation, which outperforms the existing ones with a large margin in the regime of high precision.
- Laplacian Auto-Encoders: An explicit learning of nonlinear data manifold76
This paper proposes a novel unsupervised manifold learning method termed Laplacian Auto-Encoders (LAEs), which regularizes training of auto-encoders so that the learned encoding function has the locality-preserving property for data points on the manifold.
- Geometry-Aware Self-Training for Unsupervised Domain Adaptation on Object Point Clouds75
A new method of geometry-aware self-training (GAST) for unsupervised domain adaptation of object point cloud classification, via two novel self-supervised geometric learning tasks as feature regularization, to learn a domain-shared representation of semantic categories.
- ComboVerse: Compositional 3D Assets Creation Using Spatially-Aware Diffusion Guidance73
This work presents ComboVerse, a 3D generation framework that produces high-quality 3D assets with complex compositions by learning to combine multiple models, and achieves clear improvements over existing methods in generating compositional 3D assets.
- Revisiting Realistic Test-Time Training: Sequential Inference and Adaptation by Anchored Clustering73
This work revisits TTT assumptions and categorizes TTT protocols by two key factors, and develops a realistic sequential test-time training (sTTT) protocol, under which a test-Time anchored clustering (TTAC) approach is developed to enable stronger test- time feature learning.
- Unsupervised Domain Adaptation of Black-Box Source Models68
This work proposes a simple yet effective method, termed Iterative Noisy Label Learning (IterNLL), which alternates between learning improved target models from the target subset with more reliable labels and updating the noisy target labels, and performs comparably with methods of the traditional UDA setting where the labeled source data are fully available.
- Neither Global Nor Local: Regularized Patch-Based Representation for Single Sample Per Person Face Recognition64
Experimental results show that the regularized patch-based face representation method outperforms sparse coding related face recognition methods as well as some other specially designed single sample per person face representation methods, and achieves the best performance.
- Improving the Generalization of Segmentation Foundation Model under Distribution Shift via Weakly Supervised Adaptation63
This work proposes a weakly supervised self-training architecture with anchor regularization and low-rank finetuning to improve the robustness and computation efficiency of adaptation of Segment-Anything and proves the effectiveness on 5 types of downstream segmentation tasks.
- 63
- 61
- RGBD2: Generative Scene Synthesis via Incremental View Inpainting Using RGBD Diffusion Models59
This work presents a new solution termed RGBD2 that sequentially generates novel RGBD views along a camera trajectory, and the scene geometry is simply the fusion result of these views.
- On the Robustness of Open-World Test-Time Training: Self-Training with Dynamic Prototype Expansion54
An adaptive strong OOD pruning is developed which improves the efficacy of the self- training TTT method and regularize self-training with distribution alignment and the combination yields the state-of-the-art performance on 5 OWTTT benchmarks.
- Unsupervised domain adaptation via distilled discriminative clustering54
It is interesting to observe that in the DisClusterDA framework, adding an additional loss term that explicitly learns to align class-level feature distributions across domains does harm to the adaptation performance, though more careful studies in different algorithmic frameworks are to be conducted.
- Uncertainty-Aware Clustering for Unsupervised Domain Adaptive Object Re-Identification52
A novel hierarchical clustering scheme is proposed to promote clustering quality, and an uncertainty-aware collaborative instance selection method is introduced to select images with reliable labels for model training that effectively reduces the impact of noisy labels.
- SkeletonNet: A Topology-Preserving Solution for Learning Mesh Reconstruction of Object Surfaces From RGB Images51
Two models are proposed, the Skeleton-Based Graph Convolutional Neural Network (SkeGCNN and SkeDISN), which respectively build upon and improve over the existing frameworks of explicit mesh deformation and implicit field learning for the downstream surface reconstruction task.
- Style Interleaved Learning for Generalizable Person Re-Identification50
This work proposes a novel style interleaved learning (IL) framework that not only consistently outperforms state-of-the-art methods on large-scale benchmarks for DG ReID, but also has clear advantages in computational efficiency.
- Convolutional Fine-Grained Classification With Self-Supervised Target Relation Regularization50
Inspired by recent success of the mixup style data augmentation, randomness is introduced into soft construction of dynamic target relation graphs to further explore relation diversity of target classes.
- 48
- Weakly Supervised 3D Point Cloud Segmentation via Multi-Prototype Learning47
This work designs a multi-prototype classifier, each prototype serves as the classifier weights for one subclass, and proposes two constraints respectively for updating the prototypes w.r.t. all point features and for encouraging the learning of diverse prototypes.
- Learning Parallel Dense Correspondence from Spatio-Temporal Descriptors for Efficient and Robust 4D Reconstruction45
This work presents a novel pipeline to learn a temporal evolution of the 3D human shape through spatially continuous transformation functions among cross-frame occupancy fields via explicitly learning continuous displacement vector fields from robust spatio-temporal shape representations.
- 43
- Sign-Agnostic Implicit Learning of Surface Self-Similarities for Shape Modeling and Reconstruction from Raw Point Clouds40
With a global post-optimization of local sign flipping, SAIL-S3 is able to directly model raw, un-oriented point clouds and reconstruct high-quality object surfaces and Experiments show its superiority over existing methods.
- Deep Optimized Priors for 3D Shape Modeling and Reconstruction38
This work introduces a new learning framework for 3D modeling and reconstruction that greatly improves the generalization ability of a deep generator and proposes to further optimize the learned prior and latent code according to the input physical measurements after the training.
- Exploring Diversity-Based Active Learning for 3D Object Detection in Autonomous Driving37
This work investigates diversity-based active learning (AL) as a potential solution to alleviate the annotation burden, and proposes a novel acquisition function that enforces spatial and temporal diversity in the selected samples.
- ROML: A Robust Feature Correspondence Approach for Matching Objects in A Set of Images37
This work considers simultaneously matching object instances in a set of images, where both inlier and outlier features are extracted, and proposes a novel framework, termed Robust Object Matching using Low-rank constraint (ROML), to address this problem.
- Towards Uncovering the Intrinsic Data Structures for Unsupervised Domain Adaptation Using Structurally Regularized Deep Clustering35
This work proposes a hybrid model of Structurally Regularized Deep Clustering, which integrates the regularized discriminative clustering of target data with a generative one, and term the method as H-SRDC, which outperforms all the existing methods under both the inductive and transductive settings.
- Finding Correspondence from Multiple Images via Sparse and Low-Rank Decomposition33
This work proposes a robust solution to the problem of finding the correspondence from multiple images by exploiting the priors that the rank of the ordered patterns from a set of linearly correlated images should be lower than that of the disordered patterns, and the errors among the reordered patterns are sparse.
- 32
- 31
- A practical means for calibrating an LED-based photometric stereo system31
This work presents a practical method for calibrating lighting directions based on a simple PS setup consisting of LED lamps and one camera and introduces the LED emitting model, so that the incident lighting directions for various image points can be estimated individually.
- Deep Cascade Generation on Point Sets29
This paper proposes a deep cascade network to generate 3D geometry of an object on a point cloud, consisting of a set of permutation-insensitive points, and develops a dynamically-weighted loss function for jointly penalizing the generation output of cascade levels at different training stages in a coarse-to-fine manner.
- HelixSurf: A Robust and Efficient Neural Implicit Surface Learning of Indoor Scenes with Iterative Intertwined Regularization28
A method termed Helix-shaped neural implicit Surface learning or HelixSurf, which uses the intermediate prediction from one strategy as the guidance to regularize the learning of the other one, and conducts such intertwined regularization iteratively during the learning process.
- Improving Semantic Analysis on Point Clouds via Auxiliary Supervision of Local Geometric Priors28
This article is the first attempt to propose a unique multitask geometric learning network to improve semantic analysis by auxiliary geometric learning with local shape properties, which can be either generated via physical computation from point clouds themselves as self-supervision signals or provided as privileged information.
- 28
- Revisiting Realistic Test-Time Training: Sequential Inference and Adaptation by Anchored Clustering Regularized Self-Training27
A more effective TTT approach is introduced by regularizing self-training with anchored clustering, and the improved model is referred to as TTAC++, which is demonstrated that, under all TTT protocols, TTAC++ consistently outperforms the state-of-the-art methods on five TTT datasets.
- A New Benchmark: On the Utility of Synthetic Data with Blender for Bare Supervised Learning and Downstream Domain Adaptation26
This work systematically verify the typical, important learning insights, e.g., shortcut learning, and discover the new laws of various data regimes and network architectures in generalization under the well-controlled, IID data setting enabled by 3D rendering.
- Masked Surfel Prediction for Self-Supervised Point Cloud Learning24
This work makes the first attempt to consider the local geometry information explicitly into the masked auto-encoding, and proposes a novel Masked Surfel Prediction (MaskSurf) method, which outperforms its closest competitor, Point-MAE, by 1.2\% on the real-world dataset of ScanObjectNN under the OBJ-BG setting.
- Geometry-aware Generation of Adversarial and Cooperative Point Clouds24
A novel loss function is proposed to account for imperceptible, geometry-aware deformations of point clouds, and used in an adversarial objective to attack representative models of point set classifiers, and experiments show that this method achieves stronger attacks than existing methods, without introduction of noticeable outliers and surface irregularities.
- Hallucinating multiple occluded face images of different resolutions24
A novel algorithm to super-resolve face images given multiple partially occluded inputs at different lower resolutions is proposed, which can probabilistically align multiple input images at different resolutions and recursively infer the high-resolution face image.
- Adversarial Style Augmentation for Domain Generalization23
This work introduces a novel Adversarial Style Augmentation (ASA) method, which explores broader style spaces by generating more effective statistics perturbation via adversarial training, and justifies the efficacy of AdvStyle on tasks of cross-domain classification and instance retrieval.
- Point-DAE: Denoising Autoencoders for Self-Supervised Point Cloud Learning22
This work explores a more general denoising autoencoder for point cloud learning (Point-DAE) by investigating more types of corruptions beyond masking, and identifies another effective corruption family, i.e., affine transformation.
- 21
- 20
- GUS-IR: Gaussian Splatting With Unified Shading for Inverse Rendering18
This paper starts by analyzing and comparing two prominent shading techniques popularly used for inverse rendering, forward shading and deferred shading, effectiveness in handling complex materials and proposes a unified shading solution that combines the advantages of both techniques for better decomposition.
- Vicinal and categorical domain adaptation18
This work proposes novel adversarial losses for vicinal domain adaptation (VicDA) based on CatDA, leading to Vicinal and Categorical Domain Adaptation (ViCatDA), and proposes Target Discriminative Structure Recovery (TDSR) to recover the intrinsic target discrimination damaged by adversarial feature alignment.
- 18
- 17
- Universal Domain Adaptation from Foundation Models: A Baseline Study16
This paper proposes a very simple method of target data distillation on the CLIP model, which achieves consistent improvement over the baseline across all the UniDA benchmarks, and addresses the threshold-sensitive issue encountered when using the existing H-score metric.
- STFAR: Improving Object Detection Robustness at Test-Time by Self-Training with Feature Alignment Regularization16
This work adopts a self-training paradigm to generate pseudo labeled objects with an exponential moving average model and incorporates aligning feature distributions at two output levels as regularizations to self- training for test-time adaptive object detection.
- 16
- Object Point Cloud Classification via Poly-Convolutional Architecture Search15
This paper introduces a novel concept of poly-convolutional architecture search (PolyConv search in short), which can be formulated into a Markov Decision Process problem to cast decisions for dependently selecting layer-wise aggregation kernels.
- 15
- Cumulative attribute space regression for head pose estimation and color constancy15
It is shown how the original CA space can be generalized to multiple output by the Cartesian product (CartCA), but for target spaces with more than two outputs the CartCA becomes computationally infeasible and therefore an approximate solution - multi-view CA (MvCA) - where CartCA is applied to output pairs is proposed.
- Knowledge transferred adaptive filter pruning for CNN compression and acceleration14
The normalization statistics of feature maps (NSFM) are employed as the transferable knowledge representation for soft pruning of network filters, and this method is termed as transferred adaptive pruning (TransFP), which is based on network compression techniques.
- 14
- 14
- 13
- 13
- Learning and Meshing From Deep Implicit Surface Networks Using an Efficient Implementation of Analytic Marching12
An algorithm of analytic marching is proposed, which marches among analytic cells to exactly recover the mesh captured by an implicit surface network, and it is proved that under mild conditions, the identified analytic faces are guaranteed to connect and form a closed, piecewise planar surface.
- 12
- Efficient and Context-Aware Label Propagation for Zero-/Few-Shot Training-Free Adaptation of Vision-Language Model11
This work dynamically constructs a graph over text prompts, few-shot examples, and test samples, using label propagation for inference without task-specific tuning, and introduces a context-aware feature re-weighting mechanism to improve task adaptation accuracy.
- 11
- 11
- Content-aware convolutional neural networks11
A Content-aware Convolution (CAC) is proposed that automatically detects the smooth windows and applies a 1 ×1 convolutional kernel to replace the original large kernel, and is able to effectively avoid the redundant computation on similar pixels.
- 11
- Classification of single-view object point clouds10
It is shown in this paper that performance of existing point cloud classifiers drops drastically under the considered single-view, partial setting; the phenomenon is consistent with the observation that semantic category of a partial object surface is less ambiguous only when its distribution on the whole surface is clearly specified.
- 9
- 8
- Spectral attribute learning for visual regression8
An approach to automatically discover spectral attributes which avoids manual work required for defining hand-crafted attribute representations is proposed and this spectral attribute visual regression achieves state-of-the-art accuracy in multiple applications.
- Coupling Face Registration and Super-Resolution8
A multi-resolution patch tensor based model to automatically super-resolve and register low-resolution testing face images and demonstrates the robustness of the approach on learning-based face super-resolution on real images.
- PicoPose: Progressive Pixel-to-Pixel Correspondence Learning for Novel Object Pose Estimation7
PicoPose achieves state-of-the-art performance on the seven core datasets of the BOP benchmark, demonstrating exceptional generalization to novel objects.
- 7
- Improving deep representation learning via auxiliary learnable target coding7
A margin-based triplet loss and a correlation consistency loss on the proposed target codes are designed to encourage more discriminative representations owing to enlarging between-class margins in representation space and favoring equal semantic correlation of learnable target codes respectively.
- Stochastic Consensus: Enhancing Semi-Supervised Learning with Consistency of Stochastic Classifiers7
This work proposes a new criterion based on consistency among multiple, stochastic classifiers, termed Stochastic Consensus (STOCO), and model parameters of the classifiers as a Gaussian distribution whose mean and standard deviation are jointly optimized during training.
- A semi-supervised approach for dimensionality reduction with distributional similarity7
This paper proposes a novel diffusion maps based semi-supervised algorithm for dimensionality reduction, visualization and data representation that preserves the local manifold structure in addition to separating samples in different classes, thus facilitates the classification.
- 7
- Real-Time Verification of Embodied Reasoning for Generative Skill Acquisition6
This work proposes VERGSA (Verifying Embodied Reasoning in Generative Skill Acquisition), a framework that systematically integrates real-time verification principles into embodied skill learning and establishes a seamless extension from verification of mathematical reasoning into embodied learning.
- Multi-StyleGS: Stylized Gaussian Splatting with Multiple Styles6
This paper introduces a novel 3D GS stylization solution termed Multi-StyleGS, which employs a bipartite matching mechanism to automatically identify correspondences between the style images and the local regions of the rendered images and introduces a novel semantic style loss function that employs a segmentation network to apply distinct styles to various objects of the scene.
- Semi-Supervised Unconstrained Head Pose Estimation in the Wild6
This work proposes the first semi-supervised unconstrained head pose estimation method SemiUHPE, which can leverage abundant easily available unlabeled head images and proposes dynamic entropy based filtering to adaptively remove unlabeled outliers as training progresses by updating the threshold in multiple stages.
- Understanding Attention Mechanism in Video Diffusion Models6
An in-depth perturbation analysis of the spatial and temporal attention blocks of T2V models using an information-theoretic approach indicates that temporal and spatial attention maps affect not only the timing and layout of the videos but also the complexity of spatiotemporal elements and the aesthetic quality of the synthesized videos.
- 6
- On the Adversarial Risk of Test Time Adaptation: An Investigation into Realistic Test-Time Data Poisoning6
An effective and realistic attack method is proposed that better produces poisoned samples without access to benign samples, and an effective in-distribution attack objective is derived, and two TTA-aware attack objectives are designed.
- Open-Set Semi-Supervised Learning for 3D Point Cloud Understanding6
This work proposes to selectively utilize unlabeled data through sample weighting, so that only conducive unlabeling data would be prioritized in semi-supervised learning for 3D point cloud.
- 6
- SignBot: Learning Human-to-Humanoid Sign Language Interaction5
Simulation and real-world experimental results demonstrate that SignBot can effectively facilitate human-robot interaction and perform sign language motions with diverse robots and datasets, thereby enabling natural and effective communication between robots and humans.
- Manifold-Aware Self-Training for Unsupervised Domain Adaptation on Regressing 6D Object Pose5
This paper incorporates an explicit self-supervised manifold regularization, revealing consistent cumulative target dependency across domains, to a self-training scheme to encourage more discriminative transferable representations of regression tasks.
- A Deep Learning Based Interactive Sketching System for Fashion Images Design5
This work proposes a novel bi-colored edge texture representation to synthesize textured garment images and a shading enhancer to render shading based on the grayscale edges and provides simple but effective texture cues and color constraints, so that the details can be better reconstructed.
- 5
- Towards Optimal Design of Time and Color Multiplexing Codes5
The proposed formulation and method can be readily specialized to time multiplexing, thereby making such optimized codes have a much broader application.
- Human activities recognition by head movement using partial recurrent neural network5
This paper proposes a novel use of the connectionist approach for the recognition of ten simple human activities: walking, sitting down, getting up, squatting down and standing up, in both lateral and frontal views, in an office environment.
- 4
- 4
- 3
- Exploring Human-in-the-Loop Test-Time Adaptation by Synergizing Active Learning and Model Selection3
This work first select samples for human annotation and then use the labeled data to select optimal hyper-parameters (model selection) and can always prevent choosing the worst hyper-parameters on all off-the-shelf TTA methods.
- Rapid Diffusion: Building Domain-Specific Text-to-Image Synthesizers with Fast Inference Speed3
This paper proposes Rapid Diffusion, a novel framework for training and deploying super-resolution, text-to-image latent diffusion models with rich entity knowledge injected and optimized networks and employs BladeDISC and FlashAttention techniques to optimize computational graphs of the generated models for online deployment.
- 3
- 3
- 3
- CCTV face hallucination under occlusion with motion blur3
A novel learning-based algorithm is presented that can probabilistically align multiple input images at different resolutions and recursively infer the high-resolution face image by integrating hierarchical patch-wise alignment and interframe constraints into a Bayesian framework.
- 2
- Counterfactual Supervision-Based Information Bottleneck for Out-of-Distribution Generalization2
This paper shows that the key assumption of support overlap of invariant features used in IB-IRM guarantees OOD generalization, and it is still possible to achieve the optimal solution without this assumption, and proposes a Counterfactual Supervision-based Information Bottleneck (CSIB) learning algorithm that recovers the invariants.
- 2
- 2
- ImaginationPolicy: Towards Generalizable, Precise and Reliable End-to-End Policy for Robotic Manipulation1
This paper proposes a novel Chain of Moving Oriented Keypoints (CoMOK) formulation for robotic manipulation, which uses the oriented keypoint as the action representation of a neural policy, which can be trained in an end-to-end fashion.
- BiNoMaP: Learning Category-Level Bimanual Non-Prehensile Manipulation Primitives1
This work introduces a geometry-aware post-optimization algorithm to refine bimanual hand motion trajectories from egocentric video demonstrations into executable manipulation primitives consistent with predefined motion patterns and proposes a three-stage, RL-free framework for learning structured non-prehensile skills.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Adversarial Geometric Transformations of Point Clouds for Physical Attack–
Transformation-based generation algorithms for adversarial shape and physical attack against both rotation sensitive and agnostic point classifiers are significantly outperforming existing adversarial point attackers under diverse recent defenses and the state-of-the-art physical attack methods.
- –
- –
- ShuffleMix: Improving Representations via Channel-Wise Shuffle of Interpolated Hidden States–
Compared to its direct competitor of feature augmentation -- the Manifold Mixup, the proposed ShuffleMix can gain superior generalization, owing to imposing more flexible and smooth constraints on generating samples and achieving regularization effects of channel-wise feature dropout.
- –
- –
- –
- –
- –
- Learning Reduced-Resolution and Super-Resolution Networks in Synch–
An end-to-end network (Reduced & Super-Resolution Network, RSRNet) for reduced-resolution and super-resolution without any cheap interpolation is presented, which preserves more effective information and brings better visual effect in the LR image.
- –
- –
- –
- –
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.