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
- King Ngi NganSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works70 from public data
- Thinking in Frequency: Face Forgery Detection by Mining Frequency-Aware Clues1,284
This work proposes a novel Frequency in Face Forgery Network (F3-Net), taking advantages of two different but complementary frequency-aware clues, and applies DCT as the applied frequency-domain transformation to introduce frequency into the face forgery detection.
- HydraPlus-Net: Attentive Deep Features for Pedestrian Analysis567
A new attentionbased deep neural network, named as HydraPlus-Net (HPnet), that multi-directionally feeds the multi-level attention maps to different feature layers to enrich the final feature representations for a pedestrian image.
- Morphing and Sampling Network for Dense Point Cloud Completion384
This work proposes a novel approach to complete the partial point cloud in two stages, which outperforms the existing methods in both the Earth Mover's Distance (EMD) and the Chamfer Distance (CD).
- GS3D: An Efficient 3D Object Detection Framework for Autonomous Driving361
This work leverages the off-the-shelf 2D object detector to efficiently obtain a coarse cuboid for each predicted 2D box and explores the 3D structure information of the object by employing the visual features of visible surfaces.
- CAMP: Cross-Modal Adaptive Message Passing for Text-Image Retrieval361
This paper proposes Cross-modal Adaptive Message Passing (CAMP), which adaptively controls the information flow for message passing across modalities and properly handles negative pairs and irrelevant information with an adaptive gating scheme.
- Optical Flow Guided Feature: A Fast and Robust Motion Representation for Video Action Recognition343
A novel compact motion representation for video action recognition, named Optical Flow guided Feature (OFF), which enables the network to distill temporal information through a fast and robust approach and is complementary to other motion modalities such as optical flow.
- Avatar-Net: Multi-scale Zero-Shot Style Transfer by Feature Decoration328
This paper proposes an efficient yet effective Avatar-Net that enables visually plausible multi-scale transfer for arbitrary style in real-time and demonstrates the state-of-the-art effectiveness and efficiency of the proposed method in generating high-quality stylized images.
- 261
- ForgeryNet: A Versatile Benchmark for Comprehensive Forgery Analysis234
The ForgeryNet dataset is constructed, an extremely large face forgery dataset with unified annotations in image- and video-level data across four tasks, and it is hoped that the scale, quality, and variety of this dataset will foster further research and innovation in the area of face forgeries classification, as well as spatial and temporal forgery localization etc.
- Semantics Disentangling for Text-To-Image Generation210
A novel photo-realistic text-to-image generation model that implicitly disentangles semantics to both fulfill the high- level semantic consistency and low-level semantic diversity and a visual-semantic embedding strategy by semantic-conditioned batch normalization to find diverse low- level semantics.
- DanceFormer: Music Conditioned 3D Dance Generation with Parametric Motion Transformer204
Extensive experiments demonstrate that the proposed DanceFormer, even trained by existing datasets, can generate fluent, performative, and music-matched 3D dances that surpass previous works quantitatively and qualitatively.
- Exploring Disentangled Feature Representation Beyond Face Identification163
Comprehensive evaluations demonstrate that the proposed features not only preserve state-of-the-art identity verification performance on LFW, but also acquire comparable discriminative power for face attribute recognition on CelebA and LFWA.
- Zoom-Net: Mining Deep Feature Interactions for Visual Relationship Recognition160
This work presents two new pooling cells to encourage feature interactions, and sheds light on how one could resolve ambiguous and noisy object and predicate annotations by Intra-Hierarchical trees (IH-tree).
- Improving Pedestrian Attribute Recognition With Weakly-Supervised Multi-Scale Attribute-Specific Localization154
A flexible Attribute Localization Module (ALM) is proposed to adaptively discover the most discriminative regions and learns the regional features for each attribute at multiple levels and a feature pyramid architecture is introduced to enhance the attribute-specific localization at low-levels with high-level semantic guidance.
- 144
- 140
- Back-tracing Representative Points for Voting-based 3D Object Detection in Point Clouds131
This work introduces a new 3D object detection method, named as Back-tracing Representative Points Network (BRNet), which generatively back-traces the representative points from the vote centers and also revisits complementary seed points around these generated points, so as to better capture the fine local structural features surrounding the potential objects from the raw point clouds.
- Diffusion Model is Secretly a Training-Free Open Vocabulary Semantic Segmenter115
The potential of generative text-to-image diffusion models as highly efficient open-vocabulary semantic segmenters are uncovered, and a novel training-free approach named DiffSegmenter is introduced that achieves impressive results for open-vocabulary semantic segmentation.
- Video Generation From Single Semantic Label Map111
A cVAE for predicting optical flow is employed as a beneficial intermediate step to generate a video sequence conditioned on the initial single frame, and a semantic label map is integrated into the flow prediction module to achieve major improvements in the image-to-video generation process.
- Fast-BEV: A Fast and Strong Bird’s-Eye View Perception Baseline100
This paper proposes a simple yet effective framework, termed Fast-BEV, which is capable of performing faster BEV perception on the on-vehicle chips of autonomous driving, and introduces a strong data augmentation strategy for both image and BEV space to avoid over-fitting.
- Context and Attribute Grounded Dense Captioning83
This work designs a novel end-to-end context and attribute grounded dense captioning framework consisting of a contextual visual mining module and a multi-level attribute grounded description generation module that incorporates an auxiliary supervision from hierarchical linguistic attributes to augment the distinctiveness of the learned captions.
- 70
- T2ISafety: Benchmark for Assessing Fairness, Toxicity, and Privacy in Image Generation59
This work evaluates 15 prominent diffusion models on T2ISafety and reveals several concerns including persistent issues with racial fairness, a tendency to generate toxic content, and significant variation in privacy protection across the models, even with defense methods like concept erasing.
- Toward Explainable 3D Grounded Visual Question Answering: A New Benchmark and Strong Baseline49
This work formally define and address a 3D grounded question answering (GQA) task by collecting a new 3D VQA dataset, referred to as flexible and explainable 3D GQA (FE-3DGQ a), with diverse and relatively free-form question-answer pairs, as well as dense and completely grounded bounding box annotations.
- 44
- DeepSeek: Paradigm Shifts and Technical Evolution in Large AI Models41
This paper begins by reviewing the evolution of large AI models focusing on paradigm shifts, the mainstream large language model (LLM) paradigm, and the DeepSeek paradigm, and highlights novel algorithms introduced by DeepSeek, including multi-head latent attention (MLA), mixture-of-experts (MoE), multi-token prediction (MTP), and group relative policy optimization (GRPO).
- Assessment of Multimodal Large Language Models in Alignment with Human Values39
Ch3Ef is introduced, a Compreh3ensive Evaluation dataset and strategy for assessing alignment with human expectations, and a unified evaluation strategy supporting assessment across various scenarios and different perspectives that deepen the understanding of MLLM capabilities, limitations, and the dynamic relationships between evaluation levels.
- 39
- TIGeR: Tool-Integrated Geometric Reasoning in Vision-Language Models for Robotics33
This work presents TIGeR (Tool-Integrated Geometric Reasoning), a novel framework that transforms VLMs from perceptual estimators to geometric computers by enabling them to generate and execute precise geometric computations through external tools.
- 30
- 30
- VoteHMR: Occlusion-Aware Voting Network for Robust 3D Human Mesh Recovery from Partial Point Clouds29
The first attempt to reconstruct reliable 3D human shapes from single-frame partial point clouds is made, using an end-to-end learnable method, named VoteHMR, which achieves state-of-the-art performances on two large-scale datasets, namely SURREAL and DFAUST.
- Octavius: Mitigating Task Interference in MLLMs via LoRA-MoE28
Octavius is fine-tuned in three modality setups: i.
- From GPT-4 to Gemini and Beyond: Assessing the Landscape of MLLMs on Generalizability, Trustworthiness and Causality through Four Modalities26
This paper evaluates the closed-source GPT-4 and Gemini and 6 open-source LLMs and MLLMs and uncover 14 empirical findings that are useful to understand the capabilities and limitations of both proprietary and open-source MLLMs, towards more reliable downstream multi-modal applications.
- 23
- RH20T-P: A Primitive-Level Robotic Dataset Towards Composable Generalization Agents21
This work proposes RH20T-P, a primitive-level robotic manipulation dataset, which contains about 38k video clips covering 67 diverse manipulation tasks in real-world scenarios, and standardizes a plan-execute CGA paradigm and implements an exemplar baseline called RA-P, whose positive performance on solving unseen tasks validates that the proposed dataset can offer composable generalization ability to robotic manipulation agents.
- 21
- Distortion-aware Transformer in 360° Salient Object Detection19
This paper proposes a Transformer-based model called DATFormer, which outperforms existing 2D SOD (salient object detection) and 360 SOD methods and introduces two distortion-adaptive modules to address the distortion problem.
- Powering One-Shot Topological NAS with Stabilized Share-Parameter Proxy19
The difficulties for architecture searching in such a complex space has been eliminated by the proposed stabilized share-parameter proxy, which employs Stochastic Gradient Langevin Dynamics to enable fast shared parameter sampling, so as to achieve stabilized measurement of architecture performance even in search space with complex topological structures.
- AnimaX: Animating the Inanimate in 3D with Joint Video-Pose Diffusion Models18
AnimaX is presented, a feed-forward 3D animation framework that bridges the motion priors of video diffusion models with the controllable structure of skeleton-based animation, offering a scalable solution for category-agnostic 3D animation.
- A Head Pose Tracking System Using RGB-D Camera17
A novel scheme that can train several simple predictors together, enhancing the overall prediction accuracy is proposed, and experimental results confirm its effectiveness for head movement prediction.
- Cascaded regression using landmark displacement for 3D face reconstruction16
The proposed model fitting algorithm for 3D facial expression reconstruction from a single image is robust to expression and pose changes and can reconstruct higher fidelity 3D face shape.
- 16
- 13
- 11
- 11
- SegviGen: Repurposing 3D Generative Model for Part Segmentation10
SegviGen, a framework that repurposes native 3D generative models for 3D part segmentation, leverages the structured priors encoded in pretrained 3D generative model to induce segmentation through distinctive part colorization, establishing a novel and efficient framework for part segmentation.
- 10
- 10
- 9
- Visibility Constrained Generative Model for Depth-Based 3D Facial Pose Tracking9
A statistical 3D morphable model that flexibly describes the distribution of points on the surface of the face model, with an efficient switchable online adaptation that gradually captures the identity of the tracked subject and rapidly constructs a suitable face model when the subject changes is introduced.
- 8
- 7
- 7
- Multi-Agent Amodal Completion: Direct Synthesis with Fine-Grained Semantic Guidance6
This framework uses multiple agents to collaboratively analyze occlusion relationships and determine necessary boundary expansion, yielding a precise mask for inpainting, and generates fine-grained textual descriptions, enabling Fine-Grained Semantic Guidance.
- 6
- 6
- 5
- 4
- 4
- 2
- 2
- 2
- 1
- Bags of tricks for learning depth and camera motion from monocular videos1
By combining the aforementioned tricks, this work was able to significantly improve a baseline model adapted from SfMLearner without additional inference costs.
- Spatio-Temporal Weight and Edge Adaptive De-Interlace1
A de-interlacing algorithm based on spatio-temporal weight and edge direction is presented and results show that the proposed method achieves high image quality and low hardware complexity.
- ELBO-T2IAlign: A Generic ELBO-Based Method for Calibrating Pixel-Level Text-Image Alignment in Diffusion Models–
This paper proposes ELBO-T2IAlign—a simple yet effective method to calibrate pixel-text alignment in diffusion models based on the evidence lower bound (ELBO) of likelihood, which requires no additional annotations, model retraining, or architectural modifications.
- –
- –
- Accelerating the Distribution Estimation for the Weighted Median/Mode Filters–
The proposed accelerated weighted median/mode filters are effective and efficient for a variety of applications, which have comparable performance against the current state-of-the-art counterparts and cost only a fraction of their execution time.
Publication data from OpenAlex; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-15. 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.